At Thu, 6 Jun 2002 10:14:33 +0100 (BST), Dominic Sweetman wrote:
> Difference in perception here.
> Up to and including the 2.96+ release we currently work with, no GCC
> version we've taken on has been fit for use by our MIPS customers
> without a large number of fixes, including significant changes to
> nominally non-machine-dependent code.
> If you experienced a big improvement in quality on moving to some more
> recent version, I'd love to know and that's worth telling everyone.
> But if you're saying "it's always been more or less all right" then we
> are bound to suspect you're not looking hard enough...
"I don't know." I didn't really spend a _lot_ of time staring at the
gnu toolchain until about 2 or 3 years ago (mips tools, 2 or so years
8-). Before that, i relied on tools that others had massaged ... and
invariably, yes, they did have at least a few "important" bug fixes
(often pulled in from later development versions of the tools). I
think even going back 2 and change years, we had some problems with
the versions of gcc at that time, and, for some sets of compile flags
(for us, -membedded-pic) a _lot_ of problems with binutils.
As of gcc 3.0.4 and w/ binutils 2.12.1 (with patches to each, but
generally not bug-fixes .... though we undoubtedly still have a few),
at least for us, they seem to work well for linux and for some amount
of stand-alone embedded development work.
I wouldn't disagree, BTW, that the current tools for mips seem to have
some shortcomings. I also wouldn't claim that we've comprehensively
tested the tools. 8-)
I think the goal of improving test suites to show additional bugs is a
very good one. Personally, I've been trying to make sure regression
tests get added for bugs we find & fix, but there will always be more
bugs to find.