Re: [cond-optab] Fix doubleword shifts

[Paolo Bonzini <bonzini at gnu dot org>]

> Let's see, this and the earlier "generic patches" (which
> included some patches that touched ports)

Well, those I would have posted for trunk if we weren't in stage\infty.

> plus the port-specific patch you posted...  Is that it?

Yes.  :-)

> Just hinting that maybe it's time to put the cond-optab stuff on
> a branch, particularly as your message hints that there may be
> further unseen problems worth testing and fixing - by others, to
> lighten your burden in that sense.

Well, reviews are appreciated because the ports won't change except for
relatively minor details (I now strengthened my preliminary testsuite
with long long operations, and that showed the CRIS problem by the way).
  On the other hand the devil is in those details, and I don't want to
waste people's time with the details.

When I'm sure that all (so to speak) targets build libgcc, I'll put the
stuff on a branch.  This for two reasons:

* First because it's not nice to wake up one morning and issue a call to
all target maintainers of GCC. :-)  I *do* realize the burden is mostly
on me, and I have attempted this project only after deciding that I
(should) have time to do most of the testing myself.

* Secondarily because these failures are easily fixed (also because I
have some scripts to ease debugging them), so it's not a big deal.

* Finally, because I'm not sure whether the branch will be merged
piecewise.  If so, and if there are more target independent patches to
do, I want to place them in the proper order on the branch.  I have
reasons to think that such target independent patches will cause build
failures much much more likely than runtime failures.

I'm currently building *lots* of newlibs (so far the doubleword shifts
issue the only one that surfaced).  When at least the lots of compilers
build correctly (I hope Monday) I'll branch and proceed to testing
*lots* of simulators.

Paolo