From: John Maddock (John_Maddock_at_[hidden])
Date: 2000-07-04 06:11:06
Finally (!) I think my port of regex++ to boost guidelines is ready for
review, I'm going to be computerless for the next couple of weeks, so I
guess you should consider this to be an advanced request.
The file for review is in the "regex" sub-directory of the boost vault,
note that the old "pattern_matching" directory contains code that is not
currently ready for prime-time review (Boyer-Moore and KMP code), I'll come
back to this when I have the time.
Platform test results so far:
C++ Builder 4 - OK.
C++ Builder 5 - OK
C++ Builder 5 + codeguard - OK.
VC6 - OK.
VC6 + STLPort3.2.1 - OK.
VC6 + STLPort4.0b8 - mostly OK, I had to modify STLPort to get the dll
version of the library to build (missing exports from STLPort), and
std::locale support is broken (seems like std::ctype is broken in STLPort),
I'll look at this again when the release version of STLPort 4 is available
as these problems do seem to be reported on the STLPort web site.
gcc2.9 (BeOS5) - OK
gcc 2.91 (current cygnus toolchain) - chokes over use of auto_ptr in
gcc2.95.2 (cygwin) - OK
gcc2.95.2 (mingw32) - builds OK but crashes at runtime with the notorious
"call 0" bug, for those that haven't come across this before, certain
combinations of mingw32 headers cause gcc to emit bad code (a call to a
non-existant function), if you disassemble the finished exe file you'll
find a "call 0" embedded in there instead of (usually) "call
gcc2.95.2 (mingw32 + STLPort4.0b7) - OK except that std::locale support is
broken as per VC6.
If anyone can test the library on non-win32 versions of gcc I would
appreciate the feedback, there are some fairly conservative makefiles
(egcs.mak) that will do the job, otherwise use:
for a full regression test run.
Also, I'm using the non-optimising version of VC6 at home (love the IDE
hate the compiler), so if anyone would like to re-run the regression tests
for VC6 with all optimisations on I would appreciate it.
Ditto with MWCW.
The VC6 and C++ Builder library makefiles need improving (separate
directories for Object files).
There are some user requested features that I haven't got arround to yet
(access to backrefs above \9, and some kind of save and load mechanism for
No doubt you guys will extend the list!
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk