PATCH to emit smlal/umlal for arm
Jeffrey A Law
law@cygnus.com
Tue Dec 14 21:57:00 GMT 1999
In message < 199912142150.VAA28234@cam-mail1.cambridge.arm.com >you write:
> > >> >
> > >> >
> > >> > 1999-12-10 Clinton Popetz <cpopetz@cygnus.com>
> > >> >
> > >> > * config/arm/arm.md (mulsidi3adddi, umulsidi3adddi): New patter
> ns
> > >> > for long long multiply-accumlate.
> > >>
> > >> This should have been applied to the merged-arm-thumb-backend-branch.
> > >> TARGET_ARM is not valid on the trunk, so it won't compile.
> > >
> > >Sorry. I backed this out and applied it to the branch.
> > >
> >
> > On Nick's request, I also took out the TARGET_ARM and committed this to t
> he
> > trunk.
> > -Clint
>
> How on earth did you back out the previous change? The CVS log is showing
> the head revision of arm.md as 1.40 and the date for this is today, but
> 1.40 was the head revision on Saturday when I updated then.
>
> Please don't say you've used cvs admin -o to obsolete the head revision of
> a cvs file!
That is indeed what happened. According to Clint that was SOP at his previous
job when they needed to back out an incorrect change.
Clint -- the right way to back out a change is to re-apply the patch (which
has the effect of un-doing the patch) and check in the result.
One reason cvs admin -o is considered bad is we can't recreate the sources
for a particular date. ie, consider the case where we check in a change C
on day 1. On day 5 we realize change C was bad and cvs admin -o it out of
the repository. That means we can not recreate the precise state of the
source tree for day 1 ... day 5 -- which is something we do need to do on a
fairly regular basis.
Anyway, I'll cobble up something to add to cvswrite.html so that folks won't
make this mistake in the future.
jeff
More information about the Gcc-patches
mailing list