[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[lmi] building with Visual C++
From: |
Vaclav Slavik |
Subject: |
[lmi] building with Visual C++ |
Date: |
Fri, 18 Jan 2008 22:40:49 +0100 |
User-agent: |
KMail/1.9.7 |
Hi,
I was trying to reproduce the mysterious crash in product editor
reported some time ago (with no luck, unfortunately) and because
nothing showed up under Valgrind on Linux and Valgrind isn't
available for Windows, I tried to compile LMI with Visual C++ so that
we can run the app under Purify or other checkers (none that I know
about works with Mingw).
Here's a patch I had to apply to make the code compile with Visual C++
2005 (including necessary makefiles):
http://public.vslavik.fastmail.fm/lmi.vc.patch.gz
The patch is just ugly quick hack to make it compile, it disables some
things when fixing them would be too hard (in particular, there's a
hack in place to make xml::node's private methods public and it
doesn't fool VC). Some of the changes in this patch would be trivial
to fix (e.g. code using e_mode on right side of operator / doesn't
work, but that could be fixed by adding global operator/), but some
would mean big changes -- for example, the compiler gets terribly
confused about mc_enum<> template's implementation in mc_enum.tpp,
but compiles fine if the code is moved to be inline in class' body.
Do you think it's worth having LMI buildable using VC++, should we go
on and maintain this build? (Obviously not with this patch, there
would have to be proper fixes instead of the workarounds I did.)
Regards,
Vaclav
--
PGP key: 0x465264C9, available from http://pgp.mit.edu/
- [lmi] building with Visual C++,
Vaclav Slavik <=