clang (both 3.3 and 3.4) OOM crashes on HEAD

[Kevin Oberman]

On Tue, Nov 18, 2014 at 3:23 PM, Dimitry Andric <dim at freebsd.org> wrote:

> On 18 Nov 2014, at 21:05, Dmitry Marakasov <amdmi3 at amdmi3.ru> wrote:
> >
> > * Dimitry Andric (dim at FreeBSD.org) wrote:
> >
> >>>>> This seems to be same issue as in
> >>>>> http://llvm.org/bugs/show_bug.cgi?id=20893 for which there is patch
> >>>>> review going
> >>>>> on
> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20140922/236415.html
> .
> >>>>>
> >>>>> Your test case is reproducible with the trunk of llvm/clang and
> >>>>> the patch in review resolves it. Other workaround is to disable
> >>>>> generation of debug information by removing -g flag.
> >>>>
> >>>> Hm, I had assumed this problem was fixed by importing r203311 from
> >>>> upstream llvm trunk, in head r263313.  But apparently it is not.
> >>>>
> >>>> The upstream patch seems to fix your specific test case, but it is
> still
> >>>> in review, so I prefer to wait until it is actually committed, before
> I
> >>>> import it.
> >>>
> >>> Which worries me is what we do if it's not reviewd until the release.
> We
> >>> can't just tell users to "remove -g flag", can't we?
> >>
> >> I don't expect the review to take very long, but this is how it goes
> >> with releases.  At some point, the release is cut, some bugs don't get
> >> fixed, and you will simply have to live with them.
> >>
> >> In any case, it entirely depends on how many ports it affects.  I have
> >> the impression it is just a few particular ports having this issue?
> >
> > The bug seem to have been fixed:
> >
> > http://llvm.org/bugs/show_bug.cgi?id=20893
> > http://llvm.org/bugs/show_bug.cgi?id=19031
> >
> > This is unfortunately too late for 10.1, but can we possibly have the
> > fix backported into HEAD and 10-STABLE?
>
> Yes, I have already imported the fix in r274442; I will probably get
> the MFC reminder tomorrow.
>
> I'm sorry the fix could not make it into 10.1, but apparently the first
> version of it was not the proper way of solving the problem.  The final
> version took quite a while, unfortunately.
>
> -Dimitry
>
> This seems to me like something that deserves an errata note. At the
least, it will greatly deduce the number error reports to this list and the
bug reports to bugzilla. (Well, maybe not, but I can hope.)
--
R. Kevin Oberman, Network Engineer, Retired
E-mail: rkoberman at gmail.com