Re: Binutils 2.15

[Ian Lance Taylor <ian at wasabisystems dot com>]

Alan Modra <amodra@bigpond.net.au> writes:

> Seems to me that makes mips-ecoff a candidate for deprecation and
> removal in the next, ie. 2.16, release.  The same goes for all targets
> that aren't actively maintained.  People are better served using an
> older binutils release on such targets, because the older code is more
> likely to work.

I looked a bit at the ECOFF failures.  They are not serious.  Many of
them are due to a change made by Alexandre on 2003-05-06, which
changed all the tests to pass the -32 option.  The ECOFF gas doesn't
support it, so all the tests fail.

There are also some problems because the ELF disassembler only prints
32-bit address for a 32 bit object file, but the ECOFF disassembler
prints 64-bit addresses, because there is no special sprintf_vma for
ECOFF.  The testsuite expects the 32-bit addresses.

At the moment my mips-ecoff-as is down to 22 testsuite failures, most
of which seem to be similarly bogus.  The most peculiar failures are
the empic tests; the generated code is different from what is
expected, and I'm not sure why.

Ian