This is the mail archive of the 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]

Re: sibcall.c SEGV fix

Michael Matz <> writes:

> Hi,
> On Mon, 19 Feb 2001, Zack Weinberg wrote:
> > > instead of complaining I should have fixed it.  Here it is.
> >
> > I sent a patch for this this morning.  It's effectively the same as
> Oh sorry, I haven't noticed this (I'm not on gcc-patches, and the archive
> didn't show anything related).
> > As it has, /mirable dictu/, caused me to be able to get through a
> > bootstrap for the first time in over a week, and there are no
> > differences in the failures between it and the 3.0 branch, I'm going
> Lucky you.  I also were in bootstrap land on the new-regalloc branch and
> my new version of ra.c, until HEAD was merged behind my back (I was using
> something from beginning of february).  Now I'm again staring at assembler
> output ;)

It wasn't intentionally behind your back, you'll never believe what happened.
It appears my warning from the morning never got through (still hasn't
bounced, still hasn't shown in the mailing list archives. I wonder
which puter has it), and then of course, the commit message is >200k
because it stupidly decided it  looked nicer to list every testsuite
file in consistency.vlad on a  seperate line, because of how long the
original line was (IE it wrote 

testsuite/whatever/consistency.vlad/i960-97r2: foofile.c

Of course, since there are quite a few files, we ended up with a huge
commit message (5496 lines or so, going by what GNUS tells me), and it
bounced it to me because gcc-cvs wouldn't accept it, even though *it*
generated it.

I was trying to be careful to merge the ra branch to a HEAD that would
work, so i bootstrapped with the head on linuxPPC and x86 before
merging. I had been holding off a few days since the bootstraps were
failing, as soon as i got a successful one, I sent a warning message,
waited a few hours, did the merge, waited a few more hours, and
committed it.

I can unmerge if you want me to wait a while longer, or sync to the
3.0 branch, or something.

But I also had compiled and ran make check for about 1000 tests with the merged
version, with no unexpected failures, before committing it.

So I assumed it was safe.
Sorry about that.

> Ciao,
> Michael.

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