1------------------------------------------------------------------------------ 2-- -- 3-- GNAT COMPILER COMPONENTS -- 4-- -- 5-- E R R O U T -- 6-- -- 7-- S p e c -- 8-- -- 9-- Copyright (C) 1992-2015, Free Software Foundation, Inc. -- 10-- -- 11-- GNAT is free software; you can redistribute it and/or modify it under -- 12-- terms of the GNU General Public License as published by the Free Soft- -- 13-- ware Foundation; either version 3, or (at your option) any later ver- -- 14-- sion. GNAT is distributed in the hope that it will be useful, but WITH- -- 15-- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY -- 16-- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License -- 17-- for more details. You should have received a copy of the GNU General -- 18-- Public License distributed with GNAT; see file COPYING3. If not, go to -- 19-- http://www.gnu.org/licenses for a complete copy of the license. -- 20-- -- 21-- GNAT was originally developed by the GNAT team at New York University. -- 22-- Extensive contributions were provided by Ada Core Technologies Inc. -- 23-- -- 24------------------------------------------------------------------------------ 25 26-- This package contains the routines to output error messages. They are 27-- basically system independent, however in some environments, e.g. when the 28-- parser is embedded into an editor, it may be appropriate to replace the 29-- implementation of this package. 30 31with Err_Vars; 32with Erroutc; 33with Namet; use Namet; 34with Table; 35with Types; use Types; 36with Uintp; use Uintp; 37 38with System; 39 40package Errout is 41 42 Current_Error_Source_File : Source_File_Index 43 renames Err_Vars.Current_Error_Source_File; 44 -- Id of current messages. Used to post file name when unit changes. This 45 -- is initialized to Main_Source_File at the start of a compilation, which 46 -- means that no file names will be output unless there are errors in 47 -- units other than the main unit. However, if the main unit has a pragma 48 -- Source_Reference line, then this is initialized to No_Source_File, to 49 -- force an initial reference to the real source file name. 50 51 Raise_Exception_On_Error : Nat renames Err_Vars.Raise_Exception_On_Error; 52 -- If this value is non-zero, then any attempt to generate an error 53 -- message raises the exception Error_Msg_Exception, and the error message 54 -- is not output. This is used for defending against junk resulting from 55 -- illegalities, and also for substitution of more appropriate error 56 -- messages from higher semantic levels. It is a counter so that the 57 -- increment/decrement protocol nests neatly. 58 59 Error_Msg_Exception : exception renames Err_Vars.Error_Msg_Exception; 60 -- Exception raised if Raise_Exception_On_Error is true 61 62 Warning_Doc_Switch : Boolean renames Err_Vars.Warning_Doc_Switch; 63 -- If this is set True, then the ??/?*?/?$?/?x?/?X? insertion sequences in 64 -- error messages generate appropriate tags for the output error messages. 65 -- If this switch is False, then these sequences are still recognized (for 66 -- the purposes of implementing the pattern matching in pragmas Warnings 67 -- (Off,..) and Warning_As_Pragma(...) but do not result in adding the 68 -- error message tag. The -gnatw.d switch sets this flag True, -gnatw.D 69 -- sets this flag False. 70 71 ----------------------------------- 72 -- Suppression of Error Messages -- 73 ----------------------------------- 74 75 -- In an effort to reduce the impact of redundant error messages, the 76 -- error output routines in this package normally suppress certain 77 -- classes of messages as follows: 78 79 -- 1. Identical messages placed at the same point in the text. Such 80 -- duplicate error message result for example from rescanning 81 -- sections of the text that contain lexical errors. Only one of 82 -- such a set of duplicate messages is output, and the rest are 83 -- suppressed. 84 85 -- 2. If more than one parser message is generated for a single source 86 -- line, then only the first message is output, the remaining 87 -- messages on the same line are suppressed. 88 89 -- 3. If a message is posted on a node for which a message has been 90 -- previously posted, then only the first message is retained. The 91 -- Error_Posted flag is used to detect such multiple postings. Note 92 -- that this only applies to semantic messages, since otherwise 93 -- for parser messages, this would be a special case of case 2. 94 95 -- 4. If a message is posted on a node whose Etype or Entity 96 -- fields reference entities on which an error message has 97 -- already been placed, as indicated by the Error_Posted flag 98 -- being set on these entities, then the message is suppressed. 99 100 -- 5. If a message attempts to insert an Error node, or a direct 101 -- reference to the Any_Type node, then the message is suppressed. 102 103 -- 6. Note that cases 2-5 only apply to error messages, not warning 104 -- messages. Warning messages are only suppressed for case 1, and 105 -- when they come from other than the main extended unit. 106 107 -- 7. If an error or warning references an internal name, and we have 108 -- already placed an error (not warning) message at that location, 109 -- then we assume this is cascaded junk and delete the message. 110 111 -- This normal suppression action may be overridden in cases 2-5 (but not 112 -- in case 1 or 7 by setting All_Errors mode, or by setting the special 113 -- unconditional message insertion character (!) as described below. 114 115 --------------------------------------------------------- 116 -- Error Message Text and Message Insertion Characters -- 117 --------------------------------------------------------- 118 119 -- Error message text strings are composed of lower case letters, digits 120 -- and the special characters space, comma, period, colon and semicolon, 121 -- apostrophe and parentheses. Special insertion characters can also 122 -- appear which cause the error message circuit to modify the given 123 -- string as follows: 124 125 -- Insertion character % (Percent: insert name from Names table) 126 -- The character % is replaced by the text for the name specified by 127 -- the Name_Id value stored in Error_Msg_Name_1. A blank precedes the 128 -- name if it is preceded by a non-blank character other than left 129 -- parenthesis. The name is enclosed in quotes unless manual quotation 130 -- mode is set. If the Name_Id is set to No_Name, then no insertion 131 -- occurs; if the Name_Id is set to Error_Name, then the string 132 -- <error> is inserted. A second and third % may appear in a single 133 -- message, similarly replaced by the names which are specified by the 134 -- Name_Id values stored in Error_Msg_Name_2 and Error_Msg_Name_3. The 135 -- names are decoded and cased according to the current identifier 136 -- casing mode. Note: if a unit name ending with %b or %s is passed 137 -- for this kind of insertion, this suffix is simply stripped. Use a 138 -- unit name insertion ($) to process the suffix. 139 -- 140 -- Note: the special names _xxx (xxx = Pre/Post/Invariant) are changed 141 -- to insert the string xxx'Class into the message. 142 143 -- Insertion character %% (Double percent: insert literal name) 144 -- The character sequence %% acts as described above for %, except 145 -- that the name is simply obtained with Get_Name_String and is not 146 -- decoded or cased, it is inserted literally from the names table. 147 -- A trailing %b or %s is not treated specially. 148 -- 149 -- Note: the special names _xxx (xxx = Pre/Post/Invariant) are changed 150 -- to insert the string xxx'Class into the message. 151 152 -- Insertion character $ (Dollar: insert unit name from Names table) 153 -- The character $ is treated similarly to %, except that the name is 154 -- obtained from the Unit_Name_Type value in Error_Msg_Unit_1 and 155 -- Error_Msg_Unit_2, as provided by Get_Unit_Name_String in package 156 -- Uname. Note that this name includes the postfix (spec) or (body) 157 -- strings. If this postfix is not required, use the normal % insertion 158 -- for the unit name. 159 160 -- Insertion character { (Left brace: insert file name from names table) 161 -- The character { is treated similarly to %, except that the input 162 -- value is a File_Name_Type value stored in Error_Msg_File_1 or 163 -- Error_Msg_File_2 or Error_Msg_File_3. The value is output literally, 164 -- enclosed in quotes as for %, but the case is not modified, the 165 -- insertion is the exact string stored in the names table without 166 -- adjusting the casing. 167 168 -- Insertion character * (Asterisk: insert reserved word name) 169 -- The insertion character * is treated exactly like % except that the 170 -- resulting name is cased according to the default conventions for 171 -- reserved words (see package Scans). 172 173 -- Insertion character & (Ampersand: insert name from node) 174 -- The insertion character & is treated similarly to %, except that 175 -- the name is taken from the Chars field of the given node, and may 176 -- refer to a child unit name, or a selected component. The casing is, 177 -- if possible, taken from the original source reference, which is 178 -- obtained from the Sloc field of the given node or nodes. If no Sloc 179 -- is available (happens e.g. for nodes in package Standard), then the 180 -- default case (see Scans spec) is used. The nodes to be used are 181 -- stored in Error_Msg_Node_1, Error_Msg_Node_2. No insertion occurs 182 -- for the Empty node, and the Error node results in the insertion of 183 -- the characters <error>. In addition, if the special global variable 184 -- Error_Msg_Qual_Level is non-zero, then the reference will include 185 -- up to the given number of levels of qualification, using the scope 186 -- chain. 187 -- 188 -- Note: the special names _xxx (xxx = Pre/Post/Invariant) are changed 189 -- to insert the string xxx'Class into the message. 190 191 -- Insertion character # (Pound: insert line number reference) 192 -- The character # is replaced by the string indicating the source 193 -- position stored in Error_Msg_Sloc. There are three cases: 194 -- 195 -- for package Standard: in package Standard 196 -- for locations in current file: at line nnn:ccc 197 -- for locations in other files: at filename:nnn:ccc 198 -- 199 -- By convention, the # insertion character is only used at the end of 200 -- an error message, so the above strings only appear as the last 201 -- characters of an error message. The only exceptions to this rule 202 -- are that an RM reference may follow in the form (RM .....) and a 203 -- right parenthesis may immediately follow the #. In the case of 204 -- continued messages, # can only appear at the end of a group of 205 -- continuation messages, except that \\ messages which always start 206 -- a new line end the sequence from the point of view of this rule. 207 -- The idea is that for any use of -gnatj, it will still be the case 208 -- that a location reference appears only at the end of a line. 209 210 -- Note: the output of the string "at " is suppressed if the string 211 -- " from" or " from " immediately precedes the insertion character #. 212 -- Certain messages read better with from than at. 213 214 -- Insertion character } (Right brace: insert type reference) 215 -- The character } is replaced by a string describing the type 216 -- referenced by the entity whose Id is stored in Error_Msg_Node_1. 217 -- the string gives the name or description of the type, and also 218 -- where appropriate the location of its declaration. Special cases 219 -- like "some integer type" are handled appropriately. Only one } is 220 -- allowed in a message, since there is not enough room for two (the 221 -- insertion can be quite long, including a file name). In addition, if 222 -- the special global variable Error_Msg_Qual_Level is non-zero, then 223 -- the reference will include up to the given number of levels of 224 -- qualification, using the scope chain. 225 226 -- Insertion character @ (At: insert column number reference) 227 -- The character @ is replaced by null if the RM_Column_Check mode is 228 -- off (False). If the switch is on (True), then @ is replaced by the 229 -- text string " in column nnn" where nnn is the decimal 230 -- representation of the column number stored in Error_Msg_Col plus 231 -- one (the plus one is because the number is stored 0-origin and 232 -- displayed 1-origin). 233 234 -- Insertion character ^ (Caret: insert integer value) 235 -- The character ^ is replaced by the decimal conversion of the Uint 236 -- value stored in Error_Msg_Uint_1, with a possible leading minus. 237 -- A second ^ may occur in the message, in which case it is replaced 238 -- by the decimal conversion of the Uint value in Error_Msg_Uint_2. 239 240 -- Insertion character > (Greater Than: run time name) 241 -- The character > is replaced by a string of the form (name) if 242 -- Targparm scanned out a Run_Time_Name (see package Targparm for 243 -- details). The name is enclosed in parentheses and output in mixed 244 -- case mode (upper case after any space in the name). If no run time 245 -- name is defined, this insertion character has no effect. 246 247 -- Insertion character ! (Exclamation: unconditional message) 248 -- The character ! appearing anywhere in the text of a message makes 249 -- the message unconditional which means that it is output even if it 250 -- would normally be suppressed. See section above for a description 251 -- of the cases in which messages are normally suppressed. Note that 252 -- in the case of warnings, the meaning is that the warning should not 253 -- be removed in dead code (that's the only time that the use of ! 254 -- has any effect for a warning). 255 -- 256 -- Note: the presence of ! is ignored in continuation messages (i.e. 257 -- messages starting with the \ insertion character). The effect of the 258 -- use of ! in a parent message automatically applies to all of its 259 -- continuation messages (since we clearly don't want any case in which 260 -- continuations are separated from the main message). It is allowable 261 -- to put ! in continuation messages, and the usual style is to include 262 -- it, since it makes it clear that the continuation is part of an 263 -- unconditional message. 264 265 -- Insertion character !! (Double exclamation: unconditional warning) 266 -- Normally warning messages issued in other than the main unit are 267 -- suppressed. If the message contains !! then this suppression is 268 -- avoided. This is currently used by the Compile_Time_Warning pragma 269 -- to ensure the message for a with'ed unit is output, and for warnings 270 -- on ineffective back-end inlining, which is detected in units that 271 -- contain subprograms to be inlined in the main program. It is also 272 -- used by the Compiler_Unit_Warning pragma for similar reasons. 273 274 -- Insertion character ? (Question: warning message) 275 -- The character ? appearing anywhere in a message makes the message 276 -- warning instead of a normal error message, and the text of the 277 -- message will be preceded by "warning:" in the normal case. The 278 -- handling of warnings if further controlled by the Warning_Mode 279 -- option (-w switch), see package Opt for further details, and also by 280 -- the current setting from pragma Warnings. This pragma applies only 281 -- to warnings issued from the semantic phase (not the parser), but 282 -- currently all relevant warnings are posted by the semantic phase 283 -- anyway. Messages starting with (style) are also treated as warning 284 -- messages. 285 -- 286 -- Note: when a warning message is output, the text of the message is 287 -- preceded by "warning: " in the normal case. An exception to this 288 -- rule occurs when the text of the message starts with "info: " in 289 -- which case this string is not prepended. This allows callers to 290 -- label certain warnings as informational messages, rather than as 291 -- warning messages requiring some action. 292 -- 293 -- Note: the presence of ? is ignored in continuation messages (i.e. 294 -- messages starting with the \ insertion character). The warning 295 -- status of continuations is determined only by the parent message 296 -- which is being continued. It is allowable to put ? in continuation 297 -- messages, and the usual style is to include it, since it makes it 298 -- clear that the continuation is part of a warning message, but it is 299 -- not necessary to go through any computational effort to include it. 300 -- 301 -- Note: this usage is obsolete, use ?? ?*? ?$? ?x? ?X? to specify 302 -- the string to be added when Warn_Doc_Switch is set to True. If this 303 -- switch is True, then for simple ? messages it has no effect. This 304 -- simple form is to ease transition and will be removed later. 305 306 -- Insertion character ?? (Two question marks: default warning) 307 -- Like ?, but if the flag Warn_Doc_Switch is True, adds the string 308 -- "[enabled by default]" at the end of the warning message. For 309 -- continuations, use this in each continuation message. 310 311 -- Insertion character ?x? (warning with switch) 312 -- Like ?, but if the flag Warn_Doc_Switch is True, adds the string 313 -- "[-gnatwx]" at the end of the warning message. x is a lower case 314 -- letter. For continuations, use this on each continuation message. 315 316 -- Insertion character ?X? (warning with dot switch) 317 -- Like ?, but if the flag Warn_Doc_Switch is True, adds the string 318 -- "[-gnatw.x]" at the end of the warning message. X is an upper case 319 -- letter corresponding to the lower case letter x in the message. 320 -- For continuations, use this on each continuation message. 321 322 -- Insertion character ?*? (restriction warning) 323 -- Like ?, but if the flag Warn_Doc_Switch is True, adds the string 324 -- "[restriction warning]" at the end of the warning message. For 325 -- continuations, use this on each continuation message. 326 327 -- Insertion character ?$? (elaboration information messages) 328 -- Like ?, but if the flag Warn_Doc_Switch is True, adds the string 329 -- "[-gnatel]" at the end of the info message. This is used for the 330 -- messages generated by the switch -gnatel. For continuations, use 331 -- this on each continuation message. 332 333 -- Insertion character < (Less Than: conditional warning message) 334 -- The character < appearing anywhere in a message is used for a 335 -- conditional error message. If Error_Msg_Warn is True, then the 336 -- effect is the same as ? described above, and in particular << <X< 337 -- <x< <$< <*< have the effect of ?? ?X? ?x? ?$? ?*? respectively. If 338 -- Error_Msg_Warn is False, then the < << or <X< sequence is ignored 339 -- and the message is treated as a error rather than a warning. 340 341 -- Insertion character A-Z (Upper case letter: Ada reserved word) 342 -- If two or more upper case letters appear in the message, they are 343 -- taken as an Ada reserved word, and are converted to the default 344 -- case for reserved words (see Scans package spec). Surrounding 345 -- quotes are added unless manual quotation mode is currently set. 346 -- RM and SPARK are special exceptions, they are never treated as 347 -- keywords, and just appear verbatim, with no surrounding quotes. 348 -- As a special case, 'R'M is used instead of RM (which is not treated 349 -- as a keyword) to indicate when the reference to the RM is possibly 350 -- not useful anymore, and could possibly be replaced by a comment 351 -- in the source. 352 353 -- Insertion character ` (Backquote: set manual quotation mode) 354 -- The backquote character always appears in pairs. Each backquote of 355 -- the pair is replaced by a double quote character. In addition, any 356 -- reserved keywords, or name insertions between these backquotes are 357 -- not surrounded by the usual automatic double quotes. See the 358 -- section below on manual quotation mode for further details. 359 360 -- Insertion character ' (Quote: literal character) 361 -- Precedes a character which is placed literally into the message. 362 -- Used to insert characters into messages that are one of the 363 -- insertion characters defined here. Also used for insertion of 364 -- upper case letter sequences not to be treated as keywords. 365 366 -- Insertion character \ (Backslash: continuation message) 367 -- Indicates that the message is a continuation of a message 368 -- previously posted. This is used to ensure that such groups of 369 -- messages are treated as a unit. The \ character must be the first 370 -- character of the message text. 371 372 -- Insertion character \\ (Two backslashes: continuation with new line) 373 -- This differs from \ only in -gnatjnn mode (Error_Message_Line_Length 374 -- set non-zero). This sequence forces a new line to start even when 375 -- continuations are being gathered into a single message. 376 377 -- Insertion character | (Vertical bar: non-serious error) 378 -- By default, error messages (other than warning messages) are 379 -- considered to be fatal error messages which prevent expansion or 380 -- generation of code in the presence of the -gnatQ switch. If the 381 -- insertion character | appears, the message is considered to be 382 -- non-serious, and does not cause Serious_Errors_Detected to be 383 -- incremented (so expansion is not prevented by such a msg). This 384 -- insertion character is ignored in continuation messages. 385 386 -- Insertion character ~ (Tilde: insert string) 387 -- Indicates that Error_Msg_String (1 .. Error_Msg_Strlen) is to be 388 -- inserted to replace the ~ character. The string is inserted in the 389 -- literal form it appears, without any action on special characters. 390 391 -- Insertion character [ (Left bracket: will/would be raised at run time) 392 -- This is used in messages about exceptions being raised at run-time. 393 -- If the current message is a warning message, then if the code is 394 -- executed, the exception will be raised, and [ inserts: 395 -- 396 -- will be raised at run time 397 -- 398 -- If the current message is an error message, then it is an error 399 -- because the exception would have been raised and [ inserts: 400 -- 401 -- would have been raised at run time 402 -- 403 -- Typically the message contains a < insertion which means that the 404 -- message is a warning or error depending on Error_Msg_Warn. This is 405 -- most typically used in the context of messages which are normally 406 -- warnings, but are errors in GNATprove mode, corresponding to the 407 -- permission in the definition of SPARK that allows an implementation 408 -- to reject a program as illegal if a situation arises in which the 409 -- compiler can determine that it is certain that a run-time check 410 -- would have fail if the statement was executed. 411 412 -- Insertion character ] (Right bracket: may/might be raised at run time) 413 -- This is like [ except that the insertion messages say may/might, 414 -- instead of will/would. 415 416 -- Insertion sequence "(style)" (style message) 417 -- This appears only at the start of the message (and not any of its 418 -- continuations, if any), and indicates that the message is a style 419 -- message. Style messages are also considered to be warnings, but 420 -- they do not get a tag. 421 422 -- Insertion sequence "info: " (information message) 423 -- This appears only at the start of the message (and not any of its 424 -- continuations, if any), and indicates that the message is an info 425 -- message. The message will be output with this prefix, and if there 426 -- are continuations that are not printed using the -gnatj switch they 427 -- will also have this prefix. 428 429 -- Insertion sequence "low: " or "medium: " or "high: " (check message) 430 -- This appears only at the start of the message (and not any of its 431 -- continuations, if any), and indicates that the message is a check 432 -- message. The message will be output with this prefix. Check 433 -- messages are not fatal (so are like info messages in that respect) 434 -- and are not controlled by pragma Warnings. 435 436 ----------------------------------------------------- 437 -- Global Values Used for Error Message Insertions -- 438 ----------------------------------------------------- 439 440 -- The following global variables are essentially additional parameters 441 -- passed to the error message routine for insertion sequences described 442 -- above. The reason these are passed globally is that the insertion 443 -- mechanism is essentially an untyped one in which the appropriate 444 -- variables are set depending on the specific insertion characters used. 445 446 -- Note that is mandatory that the caller ensure that global variables 447 -- are set before the Error_Msg call, otherwise the result is undefined. 448 449 Error_Msg_Col : Column_Number renames Err_Vars.Error_Msg_Col; 450 -- Column for @ insertion character in message 451 452 Error_Msg_Uint_1 : Uint renames Err_Vars.Error_Msg_Uint_1; 453 Error_Msg_Uint_2 : Uint renames Err_Vars.Error_Msg_Uint_2; 454 -- Uint values for ^ insertion characters in message 455 456 Error_Msg_Sloc : Source_Ptr renames Err_Vars.Error_Msg_Sloc; 457 -- Source location for # insertion character in message 458 459 Error_Msg_Name_1 : Name_Id renames Err_Vars.Error_Msg_Name_1; 460 Error_Msg_Name_2 : Name_Id renames Err_Vars.Error_Msg_Name_2; 461 Error_Msg_Name_3 : Name_Id renames Err_Vars.Error_Msg_Name_3; 462 -- Name_Id values for % insertion characters in message 463 464 Error_Msg_File_1 : File_Name_Type renames Err_Vars.Error_Msg_File_1; 465 Error_Msg_File_2 : File_Name_Type renames Err_Vars.Error_Msg_File_2; 466 Error_Msg_File_3 : File_Name_Type renames Err_Vars.Error_Msg_File_3; 467 -- File_Name_Type values for { insertion characters in message 468 469 Error_Msg_Unit_1 : Unit_Name_Type renames Err_Vars.Error_Msg_Unit_1; 470 Error_Msg_Unit_2 : Unit_Name_Type renames Err_Vars.Error_Msg_Unit_2; 471 -- Unit_Name_Type values for $ insertion characters in message 472 473 Error_Msg_Node_1 : Node_Id renames Err_Vars.Error_Msg_Node_1; 474 Error_Msg_Node_2 : Node_Id renames Err_Vars.Error_Msg_Node_2; 475 -- Node_Id values for & insertion characters in message 476 477 Error_Msg_Qual_Level : Nat renames Err_Vars.Error_Msg_Qual_Level; 478 -- Number of levels of qualification required for type name (see the 479 -- description of the } insertion character). Note that this value does 480 -- not get reset by any Error_Msg call, so the caller is responsible 481 -- for resetting it. 482 483 Error_Msg_Warn : Boolean renames Err_Vars.Error_Msg_Warn; 484 -- Used if current message contains a < insertion character to indicate 485 -- if the current message is a warning message. Must be set appropriately 486 -- before any call to Error_Msg_xxx with a < insertion character present. 487 -- Setting is irrelevant if no < insertion character is present. 488 489 Error_Msg_String : String renames Err_Vars.Error_Msg_String; 490 Error_Msg_Strlen : Natural renames Err_Vars.Error_Msg_Strlen; 491 -- Used if current message contains a ~ insertion character to indicate 492 -- insertion of the string Error_Msg_String (1 .. Error_Msg_Strlen). 493 494 ----------------------------------------------------- 495 -- Format of Messages and Manual Quotation Control -- 496 ----------------------------------------------------- 497 498 -- Messages are generally all in lower case, except for inserted names 499 -- and appear in one of the following three forms: 500 501 -- error: text 502 -- warning: text 503 504 -- The prefixes error and warning are supplied automatically (depending 505 -- on the use of the ? insertion character), and the call to the error 506 -- message routine supplies the text. The "error: " prefix is omitted 507 -- in brief error message formats. 508 509 -- Reserved Ada keywords in the message are in the default keyword case 510 -- (determined from the given source program), surrounded by quotation 511 -- marks. This is achieved by spelling the reserved word in upper case 512 -- letters, which is recognized as a request for insertion of quotation 513 -- marks by the error text processor. Thus for example: 514 515 -- Error_Msg_AP ("IS expected"); 516 517 -- would result in the output of one of the following: 518 519 -- error: "is" expected 520 -- error: "IS" expected 521 -- error: "Is" expected 522 523 -- the choice between these being made by looking at the casing convention 524 -- used for keywords (actually the first compilation unit keyword) in the 525 -- source file. 526 527 -- Note: a special exception is that RM is never treated as a keyword 528 -- but instead is copied literally into the message, this avoids the 529 -- need for writing 'R'M for all reference manual quotes. A similar 530 -- exception is applied to the occurrence of the string SPARK used in 531 -- error messages about the SPARK subset of Ada. 532 533 -- In the case of names, the default mode for the error text processor 534 -- is to surround the name by quotation marks automatically. The case 535 -- used for the identifier names is taken from the source program where 536 -- possible, and otherwise is the default casing convention taken from 537 -- the source file usage. 538 539 -- In some cases, better control over the placement of quote marks is 540 -- required. This is achieved using manual quotation mode. In this mode, 541 -- one or more insertion sequences is surrounded by backquote characters. 542 -- The backquote characters are output as double quote marks, and normal 543 -- automatic insertion of quotes is suppressed between the double quotes. 544 -- For example: 545 546 -- Error_Msg_AP ("`END &;` expected"); 547 548 -- generates a message like 549 550 -- error: "end Open_Scope;" expected 551 552 -- where the node specifying the name Open_Scope has been stored in 553 -- Error_Msg_Node_1 prior to the call. The great majority of error 554 -- messages operates in normal quotation mode. 555 556 -- Note: the normal automatic insertion of spaces before insertion 557 -- sequences (such as those that come from & and %) is suppressed in 558 -- manual quotation mode, so blanks, if needed as in the above example, 559 -- must be explicitly present. 560 561 ---------------------------- 562 -- Message ID Definitions -- 563 ---------------------------- 564 565 subtype Error_Msg_Id is Erroutc.Error_Msg_Id; 566 function "=" (Left, Right : Error_Msg_Id) return Boolean 567 renames Erroutc."="; 568 -- A type used to represent specific error messages. Used by the clients 569 -- of this package only in the context of the Get_Error_Id and 570 -- Change_Error_Text subprograms. 571 572 No_Error_Msg : constant Error_Msg_Id := Erroutc.No_Error_Msg; 573 -- A constant which is different from any value returned by Get_Error_Id. 574 -- Typically used by a client to indicate absense of a saved Id value. 575 576 Warning_Msg : Error_Msg_Id := No_Error_Msg; 577 -- This is set if a warning message is generated to the ID of the resulting 578 -- message. Continuation messages have no effect. It is legitimate for the 579 -- client to set this to No_Error_Msg and then test it to see if a warning 580 -- message has been issued. 581 582 procedure Delete_Warning_And_Continuations (Msg : Error_Msg_Id); 583 -- Deletes the given warning message and all its continuations. This is 584 -- typically used in conjunction with reading the value of Warning_Msg. 585 586 function Get_Msg_Id return Error_Msg_Id renames Erroutc.Get_Msg_Id; 587 -- Returns the Id of the message most recently posted using one of the 588 -- Error_Msg routines. 589 590 function Get_Location (E : Error_Msg_Id) return Source_Ptr 591 renames Erroutc.Get_Location; 592 -- Returns the flag location of the error message with the given id E 593 594 ------------------------ 595 -- List Pragmas Table -- 596 ------------------------ 597 598 -- When a pragma Page or pragma List is encountered by the parser, an 599 -- entry is made in the following table. This table is then used to 600 -- control the full listing if one is being generated. Note that the 601 -- reason we do the processing in the parser is so that we get proper 602 -- listing control even in syntax check only mode. 603 604 type List_Pragma_Type is (List_On, List_Off, Page); 605 606 type List_Pragma_Record is record 607 Ptyp : List_Pragma_Type; 608 Ploc : Source_Ptr; 609 end record; 610 611 -- Note: Ploc points to the terminating semicolon in the List_Off and Page 612 -- cases, and to the pragma keyword for List_On. In the case of a pragma 613 -- List_Off, a List_On entry is also made in the table, pointing to the 614 -- pragma keyword. This ensures that, as required, a List (Off) pragma is 615 -- listed even in list off mode. 616 617 package List_Pragmas is new Table.Table ( 618 Table_Component_Type => List_Pragma_Record, 619 Table_Index_Type => Int, 620 Table_Low_Bound => 1, 621 Table_Initial => 50, 622 Table_Increment => 200, 623 Table_Name => "List_Pragmas"); 624 625 --------------------------- 626 -- Ignore_Errors Feature -- 627 --------------------------- 628 629 -- In certain cases, notably for optional subunits, the compiler operates 630 -- in a mode where errors are to be ignored, and the whole unit is to be 631 -- considered as not present. To implement this we provide the following 632 -- flag to enable special handling, where error messages are suppressed, 633 -- but the Fatal_Error flag will still be set in the normal manner. 634 635 Ignore_Errors_Enable : Nat := 0; 636 -- Triggering switch. If non-zero, then ignore errors mode is activated. 637 -- This is a counter to allow convenient nesting of enable/disable. 638 639 ----------------------- 640 -- CODEFIX Facility -- 641 ----------------------- 642 643 -- The GPS and GNATBench IDE's have a codefix facility that allows for 644 -- automatic correction of a subset of the errors and warnings issued 645 -- by the compiler. This is done by recognizing the text of specific 646 -- messages using appropriate matching patterns. 647 648 -- The text of such messages should not be altered without coordinating 649 -- with the codefix code. All such messages are marked by a specific 650 -- style of comments, as shown by the following example: 651 652 -- Error_Msg_N -- CODEFIX 653 -- (parameters ....) 654 655 -- Any message marked with this -- CODEFIX comment should not be modified 656 -- without appropriate coordination. 657 658 ------------------------------ 659 -- Error Output Subprograms -- 660 ------------------------------ 661 662 procedure Initialize; 663 -- Initializes for output of error messages. Must be called for each 664 -- source file before using any of the other routines in the package. 665 666 procedure Finalize (Last_Call : Boolean); 667 -- Finalize processing of error message list. Includes processing for 668 -- duplicated error messages, and other similar final adjustment of the 669 -- list of error messages. Note that this procedure must be called before 670 -- calling Compilation_Errors to determine if there were any errors. It 671 -- is perfectly fine to call Finalize more than once, providing that the 672 -- parameter Last_Call is set False for every call except the last call. 673 674 -- This multiple call capability is used to do some processing that may 675 -- generate messages. Call Finalize to eliminate duplicates and remove 676 -- deleted warnings. Test for compilation errors using Compilation_Errors, 677 -- then generate some more errors/warnings, call Finalize again to make 678 -- sure that all duplicates in these new messages are dealt with, then 679 -- finally call Output_Messages to output the final list of messages. The 680 -- argument Last_Call must be set False on all calls except the last call, 681 -- and must be set True on the last call (a value of True activates some 682 -- processing that must only be done after all messages are posted). 683 684 procedure Output_Messages; 685 -- Output list of messages, including messages giving number of detected 686 -- errors and warnings. 687 688 procedure Error_Msg (Msg : String; Flag_Location : Source_Ptr); 689 -- Output a message at specified location. Can be called from the parser 690 -- or the semantic analyzer. 691 692 procedure Error_Msg_S (Msg : String); 693 -- Output a message at current scan pointer location. This routine can be 694 -- called only from the parser, since it references Scan_Ptr. 695 696 procedure Error_Msg_AP (Msg : String); 697 -- Output a message just after the previous token. This routine can be 698 -- called only from the parser, since it references Prev_Token_Ptr. 699 700 procedure Error_Msg_BC (Msg : String); 701 -- Output a message just before the current token. Note that the important 702 -- difference between this and the previous routine is that the BC case 703 -- posts a flag on the current line, whereas AP can post a flag at the 704 -- end of the preceding line. This routine can be called only from the 705 -- parser, since it references Token_Ptr. 706 707 procedure Error_Msg_SC (Msg : String); 708 -- Output a message at the start of the current token, unless we are at 709 -- the end of file, in which case we always output the message after the 710 -- last real token in the file. This routine can be called only from the 711 -- parser, since it references Token_Ptr. 712 713 procedure Error_Msg_SP (Msg : String); 714 -- Output a message at the start of the previous token. This routine can 715 -- be called only from the parser, since it references Prev_Token_Ptr. 716 717 procedure Error_Msg_N (Msg : String; N : Node_Or_Entity_Id); 718 -- Output a message at the Sloc of the given node. This routine can be 719 -- called from the parser or the semantic analyzer, although the call from 720 -- the latter is much more common (and is the most usual way of generating 721 -- error messages from the analyzer). The message text may contain a 722 -- single & insertion, which will reference the given node. The message is 723 -- suppressed if the node N already has a message posted, or if it is a 724 -- warning and N is an entity node for which warnings are suppressed. 725 726 procedure Error_Msg_F (Msg : String; N : Node_Id); 727 -- Similar to Error_Msg_N except that the message is placed on the first 728 -- node of the construct N (First_Node (N)). Note that this procedure uses 729 -- Original_Node to look at the original source tree, since that's what we 730 -- want for placing an error message flag in the right place. 731 732 procedure Error_Msg_NE 733 (Msg : String; 734 N : Node_Or_Entity_Id; 735 E : Node_Or_Entity_Id); 736 -- Output a message at the Sloc of the given node N, with an insertion of 737 -- the name from the given entity node E. This is used by the semantic 738 -- routines, where this is a common error message situation. The Msg text 739 -- will contain a & or } as usual to mark the insertion point. This 740 -- routine can be called from the parser or the analyzer. 741 742 procedure Error_Msg_FE 743 (Msg : String; 744 N : Node_Id; 745 E : Node_Or_Entity_Id); 746 -- Same as Error_Msg_NE, except that the message is placed on the first 747 -- node of the construct N (First_Node (N)). 748 749 procedure Error_Msg_NEL 750 (Msg : String; 751 N : Node_Or_Entity_Id; 752 E : Node_Or_Entity_Id; 753 Flag_Location : Source_Ptr); 754 -- Exactly the same as Error_Msg_NE, except that the flag is placed at 755 -- the specified Flag_Location instead of at Sloc (N). 756 757 procedure Error_Msg_NW 758 (Eflag : Boolean; 759 Msg : String; 760 N : Node_Or_Entity_Id); 761 -- This routine is used for posting a message conditionally. The message 762 -- is posted (with the same effect as Error_Msg_N (Msg, N) if and only 763 -- if Eflag is True and if the node N is within the main extended source 764 -- unit and comes from source. Typically this is a warning mode flag. 765 -- This routine can only be called during semantic analysis. It may not 766 -- be called during parsing. 767 768 procedure Change_Error_Text (Error_Id : Error_Msg_Id; New_Msg : String); 769 -- The error message text of the message identified by Id is replaced by 770 -- the given text. This text may contain insertion characters in the 771 -- usual manner, and need not be the same length as the original text. 772 773 function First_Node (C : Node_Id) return Node_Id; 774 -- Given a construct C, finds the first node in the construct, i.e. the one 775 -- with the lowest Sloc value. This is useful in placing error msgs. Note 776 -- that this procedure uses Original_Node to look at the original source 777 -- tree, since that's what we want for placing an error message flag in 778 -- the right place. 779 780 function First_Sloc (N : Node_Id) return Source_Ptr; 781 -- Given the node for an expression, return a source pointer value that 782 -- points to the start of the first token in the expression. In the case 783 -- where the expression is parenthesized, an attempt is made to include 784 -- the parentheses (i.e. to return the location of the initial paren). 785 786 function Get_Ignore_Errors return Boolean; 787 -- Return True if all error calls are ignored. 788 789 procedure Purge_Messages (From : Source_Ptr; To : Source_Ptr) 790 renames Erroutc.Purge_Messages; 791 -- All error messages whose location is in the range From .. To (not 792 -- including the end points) will be deleted from the error listing. 793 794 procedure Remove_Warning_Messages (N : Node_Id); 795 -- Remove any warning messages corresponding to the Sloc of N or any 796 -- of its descendent nodes. No effect if no such warnings. Note that 797 -- style messages (identified by the fact that they start with "(style)") 798 -- are not removed by this call. Basically the idea behind this procedure 799 -- is to remove warnings about execution conditions from known dead code. 800 801 procedure Remove_Warning_Messages (L : List_Id); 802 -- Remove warnings on all elements of a list (Calls Remove_Warning_Messages 803 -- on each element of the list, see above). 804 805 procedure Set_Ignore_Errors (To : Boolean); 806 -- Following a call to this procedure with To=True, all error calls are 807 -- ignored. A call with To=False restores the default treatment in which 808 -- error calls are treated as usual (and as described in this spec). 809 810 procedure Set_Warnings_Mode_Off (Loc : Source_Ptr; Reason : String_Id) 811 renames Erroutc.Set_Warnings_Mode_Off; 812 -- Called in response to a pragma Warnings (Off) to record the source 813 -- location from which warnings are to be turned off. Reason is the 814 -- Reason from the pragma, or the null string if none is given. 815 816 procedure Set_Warnings_Mode_On (Loc : Source_Ptr) 817 renames Erroutc.Set_Warnings_Mode_On; 818 -- Called in response to a pragma Warnings (On) to record the source 819 -- location from which warnings are to be turned back on. 820 821 procedure Set_Specific_Warning_Off 822 (Loc : Source_Ptr; 823 Msg : String; 824 Reason : String_Id; 825 Config : Boolean; 826 Used : Boolean := False) 827 renames Erroutc.Set_Specific_Warning_Off; 828 -- This is called in response to the two argument form of pragma Warnings 829 -- where the first argument is OFF, and the second argument is a string 830 -- which identifies a specific warning to be suppressed. The first argument 831 -- is the start of the suppression range, and the second argument is the 832 -- string from the pragma. Loc is the location of the pragma (which is the 833 -- start of the range to suppress). Reason is the reason string from the 834 -- pragma, or the null string if no reason is given. Config is True for the 835 -- configuration pragma case (where there is no requirement for a matching 836 -- OFF pragma). Used is set True to disable the check that the warning 837 -- actually has the effect of suppressing a warning. 838 839 procedure Set_Specific_Warning_On 840 (Loc : Source_Ptr; 841 Msg : String; 842 Err : out Boolean) 843 renames Erroutc.Set_Specific_Warning_On; 844 -- This is called in response to the two argument form of pragma Warnings 845 -- where the first argument is ON, and the second argument is the prefix 846 -- of a specific warning to be suppressed. The first argument is the end 847 -- of the suppression range, and the second argument is the string from 848 -- the pragma. Err is set to True on return to report the error of no 849 -- matching Warnings Off pragma preceding this one. 850 851 function Compilation_Errors return Boolean; 852 -- Returns True if errors have been detected, or warnings in -gnatwe (treat 853 -- warnings as errors) mode. Note that it is mandatory to call Finalize 854 -- before calling this routine. Always returns False in formal verification 855 -- mode, because errors issued when analyzing code are not compilation 856 -- errors, and should not result in exiting with an error status. 857 858 procedure Error_Msg_CRT (Feature : String; N : Node_Id); 859 -- Posts a non-fatal message on node N saying that the feature identified 860 -- by the Feature argument is not supported in either configurable 861 -- run-time mode or no run-time mode (as appropriate). In the former case, 862 -- the name of the library is output if available. 863 864 procedure Error_Msg_PT (E : Entity_Id; Iface_Prim : Entity_Id); 865 -- Posts an error on protected type entry or subprogram E (referencing its 866 -- overridden interface primitive Iface_Prim) indicating wrong mode of the 867 -- first formal (RM 9.4(11.9/3)). 868 869 procedure Error_Msg_Ada_2012_Feature (Feature : String; Loc : Source_Ptr); 870 -- If not operating in Ada 2012 mode, posts errors complaining that Feature 871 -- is only supported in Ada 2012, with appropriate suggestions to fix this. 872 -- Loc is the location at which the flag is to be posted. Feature, which 873 -- appears at the start of the first generated message, may contain error 874 -- message insertion characters in the normal manner, and in particular 875 -- may start with | to flag a non-serious error. 876 877 procedure dmsg (Id : Error_Msg_Id) renames Erroutc.dmsg; 878 -- Debugging routine to dump an error message 879 880 ------------------------------------ 881 -- SPARK Error Output Subprograms -- 882 ------------------------------------ 883 884 -- The following routines are intended to report semantic errors in SPARK 885 -- constructs subject to aspect/pragma SPARK_Mode. Note that syntax errors 886 -- must be reported using the Error_Msg_XXX routines. This allows for the 887 -- partial analysis of SPARK features when they are disabled via SPARK_Mode 888 -- set to "off". 889 890 procedure SPARK_Msg_N (Msg : String; N : Node_Or_Entity_Id); 891 pragma Inline (SPARK_Msg_N); 892 -- Same as Error_Msg_N, but the error is suppressed if SPARK_Mode is Off. 893 -- The routine is inlined because it acts as a simple wrapper. 894 895 procedure SPARK_Msg_NE 896 (Msg : String; 897 N : Node_Or_Entity_Id; 898 E : Node_Or_Entity_Id); 899 pragma Inline (SPARK_Msg_NE); 900 -- Same as Error_Msg_NE, but the error is suppressed if SPARK_Mode is Off. 901 -- The routine is inlined because it acts as a simple wrapper. 902 903 ------------------------------------------ 904 -- Utility Interface for Casing Control -- 905 ------------------------------------------ 906 907 procedure Adjust_Name_Case (Loc : Source_Ptr); 908 -- Given a name stored in Name_Buffer (1 .. Name_Len), set proper casing. 909 -- Loc is an associated source position, if we can find a match between 910 -- the name in Name_Buffer and the name at that source location, we copy 911 -- the casing from the source, otherwise we set appropriate default casing. 912 913 procedure Set_Identifier_Casing 914 (Identifier_Name : System.Address; 915 File_Name : System.Address); 916 -- This subprogram can be used by the back end for the purposes of 917 -- concocting error messages that are not output via Errout, e.g. 918 -- the messages generated by the gcc back end. 919 -- 920 -- The identifier is a null terminated string that represents the name of 921 -- an identifier appearing in the source program. File_Name is a null 922 -- terminated string giving the corresponding file name for the identifier 923 -- as obtained from the front end by the use of Full_Debug_Name to the 924 -- source file referenced by the corresponding source location value. On 925 -- return, the name is in Name_Buffer, null terminated with Name_Len set. 926 -- This name is the identifier name as passed, cased according to the 927 -- default identifier casing for the given file. 928 929end Errout; 930