This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: state of 3.2.1-pre: how far from release?


David Edelsohn <dje@watson.ibm.com> writes:

> >>>>> Joe Buck writes:
> 
> Joe> bootstrap/8362
> Joe> This is PowerPC specific and appears to be a blocker.
> 
> Joe> 3.2.1 has a vast number of bug fixes over 3.2.  If we can get 8362
> Joe> fixed, I'd say ship the sucker, it is in the best interest of the
> Joe> users to get the fixes out there (especially all the fixes to the
> Joe> x86 backend).  The other bugs are not that important.
> 
> 	I can guard the source code in abi_check.cc which elicits the ICE
> so that the compiler will bootstrap.  This will not fix the bug.  The
> testcase takes a very long time to compile when running the compiler in
> the debugger, so it is difficult to debug.  The rest of libstdc++-v3
> compiles, so this bug is not extremely common but it is not complicated
> either.
> 
> 	I believe the testcase is uncovering a serious, latent bug in the
> common part of the compiler.  GCC has plenty of other "how did that ever
> work" bugs, so GCC 3.2.1 could be released with this bug too.
> 
> 	Without help from GCC experts who understand the reload pass, this
> will take a while to debug and fix.

In the interest of getting this release shipped, I would recommend
disabling the load_multiple pattern on the branch.  Whatever fix is
finally produced, it will touch reload/reg-alloc and so probably won't
be suitable for the branch at this point before a release anyway.

-- 
- Geoffrey Keating <geoffk@geoffk.org>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]