c++ frontend bugs

I've encountered a few bugs including ICE and reject-valid when compiling a program with llvmg++ which compiles fine with standalone g++.

Should I report those to your bugzilla or is it likely that these bugs are gcc bugs in the version of gcc you're using?
if so, are there any plans to port the frontend to a newer gcc? (gcc 4.0 will be out in a few weeks)

the versions:
llvm 1.4 from the debian package that got added to the "testing" distribution today
gcc version 3.4-llvm 20030924 (experimental)
gcc-Version 3.4.4 20050314 (prerelease) (Debian 3.4.3-12)
gcc version 4.0.0 20050418 (prerelease)

I can check this myself if you provide me with the exact version of gcc llvm is based on(and it's still available).

Thanks,

Markus F.X.J. Oberhumer schrieb:

Stefan Strasser wrote:

I've encountered a few bugs including ICE and reject-valid when compiling a program with llvmg++ which compiles fine with standalone g++.

Should I report those to your bugzilla or is it likely that these bugs are gcc bugs in the version of gcc you're using?
if so, are there any plans to port the frontend to a newer gcc? (gcc 4.0 will be out in a few weeks)

the versions:
llvm 1.4 from the debian package that got added to the "testing" distribution today
gcc version 3.4-llvm 20030924 (experimental)
gcc-Version 3.4.4 20050314 (prerelease) (Debian 3.4.3-12)
gcc version 4.0.0 20050418 (prerelease)

I can check this myself if you provide me with the exact version of gcc llvm is based on(and it's still available).

I have already opened a "wishlist" bug to merge llvm-gcc with current gcc 3.4 CVS branch - http://llvm.cs.uiuc.edu/bugs/show_bug.cgi?id=498 .

Actually I don't think it would be _too_ much work, but it requires intimate knowledge of the llvm internals (and a good merge tool), so this is probably outside the scope of an external contributor.

by "check this myself" I was talking about checking, if this is a gcc or a llvm bug by compiling and testing the outdated standalone gcc, not merging current gcc.
cfrontend with EMIT_LLVM defined to 0 doesn't compile so I can't test with this source.

I've encountered a few bugs including ICE and reject-valid when compiling a program with llvmg++ which compiles fine with standalone g++.

Ok.

Should I report those to your bugzilla or is it likely that these bugs are gcc bugs in the version of gcc you're using?

There have been a large number of bug fixes to the LLVM C++ frontend since 1.4. Probably the easiest way to check to see if it's already fixed is to see if you can repro the problem with the demo page (http://llvm.cs.uiuc.edu/demo).

If the bug still exists, please file a report! :slight_smile:

if so, are there any plans to port the frontend to a newer gcc? (gcc 4.0 will be out in a few weeks)

There are not any concrete plans at present, but it is something that I would like to do at some point.

-Chris

Markus F.X.J. Oberhumer schrieb:

The reason behind this is that llvm-gcc is based on a pre-3.4 snapshot ("3.4-llvm 20030924")

then it's not unlikely that these are llvm bugs since my code also compiles with gcc 3.3
filed PR 551 and 552

That's not entirely certain. GCC 3.4 includes a completely rewritten C++ parser, so bugs in 3.4 that are not in 3.3 is likely. I'll take a look at the PR's in the next couple of days, thanks for filing them!

-Chris