copyright update |
|
More...
|
over 13 years ago
|
handle 1 bit bitmaps with transparency |
|
More...
|
over 13 years ago
|
ppmquant location |
|
More...
|
over 16 years ago
|
another LE32 fix |
|
More...
|
about 17 years ago
|
LE32 fix |
|
More...
|
about 17 years ago
|
win compile |
|
More...
|
about 17 years ago
|
Add notes about string concatenation, HEX + -H considerations, "UL" vs "LU". Also fix markup typo. |
|
More...
|
over 18 years ago
|
Unify .s and .ss strings: string concatenation is now available everywhere. Document concatenation via both "+" and "\". |
|
More...
|
over 18 years ago
|
Fixed markup typos, as detected by the W3C validator (and added an HTML 4.01 Transitional doctype). The only remaining complaint from the validator is about our nasty habit of using <ul>...</ul> (without any intervening <li> tags) as a substitute for indentation. |
|
More...
|
over 18 years ago
|
ParseNavigationList() and ParseNavigationMap() are given a p_int, not plain int. Where p_int is larger than int (e.g. AMD64), this was previously not entirely initialising RCNAVIGATION.numOfObjects, leading to a crash exhibited by e.g. the fnav_test.rcp sample. |
|
More...
|
over 18 years ago
|
Use the half-arsed vaguely right <stdint.h>-style typedefs throughout these structs; in particular, use PILRC_ULONG rather than plain "long", as the latter is wrong on 64-bit platforms. The change in signedness is harmless -- bi[XY]PelsPerMeter are unused, and the only usages of bi{Width,Height} are in fact to pass them to a function expecting an unsigned. |
|
More...
|
over 18 years ago
|
Remember to use p_int (not just int) in a struct that's going to be given to CbEmitStruct(). [Fix #1 for 1244602; thanks to Lucas Wall] |
|
More...
|
over 18 years ago
|
Don't process #defines (in .rcp files) when they are in text that is being skipped due to #ifdefs. |
|
More...
|
about 19 years ago
|
Diagnostic() should treat POSITION->szFilename==NULL as a non-localised diagnostic as well. Also print "pilrc: " (and perhaps "warning: ") in front of such diagnostics. |
|
More...
|
about 19 years ago
|
Some time ago I made Usage() follow the GNU Coding Standards for usage displays w.r.t. printing to stdout and returning EXIT_SUCCESS. However Usage() is really misnamed -- it is mostly used to exit in the case of command line option processing errors, and in this case returning EXIT_SUCCESS is a bug. |
|
More...
|
about 19 years ago
|
Add proper implementations of FIsString() and PchGetString(). These allow concatenation involving both simple string tokens and string-valued #defines, and go to some effort to provide useful error messages. |
|
More...
|
about 19 years ago
|
Update copyright year (as printed) to 2005. |
|
More...
|
about 19 years ago
|
The non-trivial changes to the various parsers so that they use the new FIsString() and PchGetString() string parser functions. |
|
More...
|
about 19 years ago
|
Move UngetTok()'s token buffer into INPUTCONTEXT. When you open a new input stream, any ungotten tokens from the previous stream should not appear immediately, but rather after the new input stream has run out and input has reverted to the previous stream. |
|
More...
|
about 19 years ago
|
Initialise the lexer in OpenInputFile(), i.e., each time a new input stream is opened. |
|
More...
|
about 19 years ago
|
FIsString() and PchGetString() are new functions encapsulating string parsing. A later checkin will provide proper implementations and remove the old PchGetSz/PchGetSzMultiLine functions. This checkin simply provides dummy versions and the trivial renaming parts of the changes needed to use them. |
|
More...
|
about 19 years ago
|
* Use DesirableLocale() instead of ad hoc szLocaleP checking. * Where practical, move locale tests to after we've finished parsing, so that a buggy string skipping early return is not needed. |
|
More...
|
about 19 years ago
|
Rename FPeekTok() to PeekTok() -- "F" is inappropriate because this function doesn't return a boolean flag. |
|
More...
|
about 19 years ago
|
Check DesirableLocale() only *after* all parsing is complete. Otherwise if the object is not to be emitted then parts of its source code will be left behind in the input stream, causing spurious syntax errors. |
|
More...
|
about 19 years ago
|
Open and close the .ro temporary file before we start appending resource data to it, so that it exists even if there are no resources to be added to it. (This is to be sure an unusual situation that is probably erroneous in itself, but "<tempfile> not found" is not a very friendly way to diagnose it.) |
|
More...
|
about 19 years ago
|
Exterminate ObjectDesiredInOutputLocale() in favour of the more flexible and comprehensible DesirableLocale(). |
|
More...
|
about 19 years ago
|
Fix bug in DesirableLocale(): If no -Loc option has been given, then objects with locales should NOT be emitted. Particularly: if you have localised versions of the same form for several locales, you don't want to emit all of them! |
|
More...
|
about 19 years ago
|
According to PalmSource's FileFormats.pdf (p7), the database name must be NUL-terminated within its 32 bytes and is thus a C string of maximum length 31, rather than a 32 character fixed-length string to be manipulated with strncpy()/strncat()/etc. |
|
More...
|
about 19 years ago
|
Update copyright date; this code changed on 2005-01-19. |
|
More...
|
about 19 years ago
|
Split out INPUTCONTEXT members relating to the file itself into a new FILELINE structure. Diagnostic() now takes a FILELINE rather than just filename & lineno, thus providing everything needed by the CW plugin. |
|
More...
|
over 19 years ago
|