Re: [Gm2] Re: Solaris GM2 problem

[Gaius Mulley]

john o goyo <address@hidden> writes:

> Greetings:
> 
> Gaius Mulley wrote:
> > john o goyo <address@hidden> writes:
> > make gm2/gm2-compiler-paranoid/SymbolTable.o
> Success!

> > make gm2/gm2-compiler-paranoid/SymbolKey.o
> Success!

> > make gm2/gm2-compiler-paranoid/Lists.o
> Success!

> > make gm2/gm2-compiler-libs/ASCII.o
> > make gm2/gm2-compiler-libs/FIO.o

> "gm2-compiler-libs"?  The distribution has gm2-libs, gm2-libs-boot,
> gm2-libs-paranoid, and gm2-libs-iso.  The first two both have these
> object modules.

oops, sorry it should have read:

make gm2/gm2-libs/ASCII.o
make gm2/gm2-libs/FIO.o

> > ASCII should be the easiest to compile, in the past I've also seen
> > the
> > others fail in similar circumstances. Also do you know if the
> > regression tests fail - in particular the "run" tests?
> 
> I shall mail you the results separately (so as not to clog up the list).
> In summary:
> 
>       # of expected passes            2148
>       # of unexpected failures        24

thanks for these results. This is encouraging as I get exactly the
same results. However the regression test is testing stage1/gm2/cc1gm2
and therefore the compile only tests are of limited value, but the run
tests are of much greater value and they all pass - which is very
encouraging.

So it would be interesting (and relatively easy) to:

cd <build-dir>
cp stage2/gm2/cc1gm2 stage1/gm2/cc1gm2
make check-gm2

which has the effect of running the regression tests using
stage2/gm2/cc1gm2

[Probably sensible to remove the build directory and start
the build all over again, after cheating with the copy above]

Anyhow it looks as if FifoQueue has exposed a bug which does not
manifest itself with other pointer related modules (Lists, SymbolKey).
I think we need to run the compilation of FifoQueue (using
stage2/gm2/cc1gm2) under gdb and see whereabouts it is going wrong.
In effect we need to see whereabouts the flow of control differs
between the original stage1/gm2/cc1gm2 and stage2/gm2/cc1gm2.

regards,
Gaius