1 BUGS 2 3 4This is a place where I can record bugs and the like as I spot them. I HOPE 5that as they get resolved I will either remove them from the list here or 6write comments about the fixes. But reality suggests that that will not 7always happen! I will put a "**" in front of ones I believe are fixed. 8 9To start with these are mainly going to be not just CSL-specific but 10issues within CSL itself. Some cleverer way to organise bugs than a simple 11text file is probably called for. 12 13**25 May 2008 (CSL) 14 On Windows (at least I tested on Vista-64 but with a 32-bit build) 15 launch csl or fwindemo (etc) then try the Font/Font menu. When the 16 application is compiled normally that works as expected. But when 17 both FOX and the application are compiled with debugging options 18 it appears to loop. 19 29 May: --enable-debug but -O0 for the compilation of FOX makes it behave 20 OK again. Some time I will need to compile the files that make up FOX one 21 by one with -O1 and see which one leads to that change. Meanwhile 22 adjustments to Xft font support lead to the X11 version responding to 23 the FONT menu by only showing the Computer Modern family of fonts 24 available, and that is a bit of a disaster since none of those are 25 fixed pitch. 26 31 May: Font/Font better on Windows, but on Linux it does not display 27 the current font as its default selection in the expected manner. 28 2 June: changes to FXTerminal and FXFontSelector fix that. 29 30**25 May 2008 (CSL) 31 The "^T" keystroke exchanges the characters, but in the windowed 32 interfaces the flipped ones are redrawn in black not the normal 33 red input colour. The input line gets redrawn when some other character 34 is entered. 27 May: this was true for a range of other editing directives 35 and may now be fixed by using eg replaceStyledText where I used to use 36 just replaceText in FXTerminal.cpp 37 38**25 May 2008 (CSL) 39 The "^L" keystroke clears the screen, but typing a character afterwards 40 led to a "bad argument to changeStyle" diagnostic because I had not 41 reset promptEnd. Should now be fixed, but ought I to re-insert the current 42 prompt on clearing the screen? 43 44**26 May 2008 (CSL) 45 termdemo on Linux behaves OK, but "fwindemo -w" or fwindemo without 46 DISPLAY set seems to fail to use colours and fails to accept control 47 keys when I tested it from an xterm using remote access. 28 May fixed - the 48 build within FOX had not detected (n)curses etc properly. 49 50**26 May 2008 (CSL) 51 "fontdemo cmmi10" behaved OK from a remote xterm (but PERHAPS the font was 52 present on the x-term?) but showmathdemo crashed with a segmentation fault! 53 Ditto when run native. 31 May: Rework of the use of Xft etc fixes this. 54 5505 June 2008 (CSL, ?PSL) 56 Redlog does not run its test programs, so I have commented it out from the 57 list of things to test! 58 5909 July 2008 (CSL) 60 Xft/fontconfig startup on MacOS can be VERY slow at least the first time, 61 to the extent that it seems that the programs have just got stuck. At 62 present by belief is that this is merely (!) that it can take minutes 63 of CPU time to set up the fontconfig caches by inspecting all the fonts. 64 6509 July 2008 (CSL) 66 Windows64 (via mingw64) behaves sort of OK for CSL, but fontdemo and 67 showmathdemo just display an empty frame. Hmmm. 68 6909 July 2008 (PSL) 70 The sets of pre-build PSL "bpsl" etc files need review by somebody who 71 understands them, and the scripts/pslver.sh script reviewed so it picks 72 the right one. I am uncertain about the "-td" parameter value that will 73 be best, and the way that the varies between 32 and 64-bit models. 74 7509 July 2008 (CSL, PSL) 76 A scheme to re-build individual packages (and to collect relevant 77 dependency information) is needed. 78 7909 July 2008 (PSL) 80 The build-sequence needs integrating into Makefile.am and it then needs 81 dependency information so it will be re-done in relevant situations. Also 82 the destination for all the log files needs to be reviewed. 83 8409 July 2008 (CSL, PSL) 85 Perhaps a scheme to run individual tests is needed. 86 87 8809 July 2008 (PSL) 89 The win32/bpsl.exe is a hacked one that uses static allocation and hence 90 survives Vista-64, but it is probably inflexible when that trick is not 91 necessary 92 9309 July 2008 (PSL) 94 script to launch REDUCE needed to go in "bin" directory 95 9609 July 2008 (PSL) 97 On Windows, how should I select between bpsl.exe and bpslw.exe? In 98 general if people want to use redfront or emacs how can that be done? 99 10009 July 2008 (redfront) 101 Needs altering from GNU readline to the BSD-licensed "libedit" 102 (see http://sourceforge.net/projects/libedit (probably defunt) and 103 http://www.thrysoee.dk/editline/ which is an autoconfed variant 104 on what you find at 105 cvsweb.netbsd.org/bsdweb.cgi/src/lib/libedit/?sortby=date#dirlist). 106 This aims for total compatibility with GNU readline but with a license 107 that makes redistribution of binaries less complicated and that will 108 hence fit with this project better. 109 11009 July 2008 (PSL) 111 A quick test seemed to show that the amd64 build on Linux took a LOT 112 longer than the linux(32) build on the same machine. Why? Is this to do 113 with how much memory was allocated? 114 11509 July 2008 (CSL) 116 Should I put "images" as directories so that "make" can observe 117 date-stamps on individual fasl files directly and easily? 118 11924 July 2008 (CSL) 120 Much broken while I am in the middle of changing so that image files can 121 be replaced by image directories! 122 12324 July 2008 (CSL) 124 The build "--with-cygwin" appears not to link at present. Not making 125 investigation an especially high priority. 126 127 12825 July 2008 (PSL mostly I think) 129 support/remake.red and PERHAPS other places express views about where 130 files live, as in "$reduce/lisp/psl/$MACHINE/red/" in update!-fasl2 131 and maybe other places. This does not match the new file layout and is 132 a bit delicate anyway. 133 134=============================================================================== 135