[Gm2] Re: gm2 Cygwin port

[Gaius Mulley]

"Gerrit P. Haase" <address@hidden> writes:

> Hello Gaius,
> 
> 
> At 7. June 2004 20:39 you wrote:
> > Has gm2 come to life under cygwin yet?
> 
> 
> The bootstrap runs nearly through, but I got an error in the final
> compare: 
> 
> Bootstrap comparison failure!
> gm2/mod_init.o differs
> make[1]: *** [compare] Error 1
> 
> I guess, the Make-lang.in is simply not designed for a usual three
> stage gcc bootstrap.
> 
> I really want to include gm2 into the Cygwin release, but I need to
> use the bootstrap target. 
> 
> I fixed it now with a kludge, because I think the relevant mod_init.o
> isn't important for the compiler, and don't need to be compared, so I
> just remove it everytime after it was used.

yes a mod_init.o is created every time we link a gm2 tool: gm2l, gm2m,
cc1gm2 etc, maybe they should have distinct filenames. This would
fix the problem I guess..

> This seems to work.  There were some more changes needed, please see
> the attached patchfile for all the changes I included, some are just
> kludges, some are bugfixes, a workaround is needed for the parts where
> you're checking for a.out files which are named a.exe on
> Cygwin/Windows.  Some more files need a rename:
> $ tar xjf ../gcc-3.3.3/gcc-gm2-3.3.3.tar.bz2
> tar: gcc-3.3.3/gcc/gm2/gm2-compiler/m2lex.def: Cannot open: No such file or 
> directory
> tar: gcc-3.3.3/gcc/gm2/m2/comp/libs/Math.def: Cannot open: No such file or 
> directory
> tar: gcc-3.3.3/gcc/gm2/m2/comp/libs/Math.mod: Cannot open: No such file or 
> directory
> 
> So there are/were M2lex.def & m2lex.def, math.def & Math.def and
> math.mod & Math.mod.

ok I'll look into this asap..


> My Mage-Lang.in patch is still wrong, but it works finally to use the
> bootstrap, I'm getting now just few errors during the tests:
>                 === gm2 tests ===
> 
> Schedule of variations:
>     unix
> 
> Running target unix

[.. snip .. ]

> Running /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/pim/pass/gm2.exp ...
> FAIL: gm2/pim/pass/procmod.mod,  -g  
> FAIL: gm2/pim/pass/procmod.mod,  -O  
> FAIL: gm2/pim/pass/procmod.mod,  -g -O  
> Running /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/pim/pass/gm2.exp ...
> FAIL: gm2/pim/pass/procmod.mod,  -g  
> FAIL: gm2/pim/pass/procmod.mod,  -O  
> FAIL: gm2/pim/pass/procmod.mod,  -g -O  
> Running /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/pim/run/fail/gm2.exp ...
> ERROR: tcl error sourcing 
> /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/pim/run/fail/gm2.exp.
> ERROR: can't read "option_list": no such variable
>     while executing
> "foreach option $option_list {

[ ..snip ..]

> Running /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/pim/run/fail/gm2.exp ...
> Running /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/ulmlib/pass/gm2.exp ...
> Running /gcc/gcc-3.3.3/gcc-3.3.3-1/gcc/testsuite/gm2/ulmlib/pass/gm2.exp ...
> 
>                 === gm2 Summary ===
> 
> # of expected passes            3048
> # of unexpected failures        6

Well done! All these errors I get under GNU/Linux anyway. The last
error (run test) is a bug in the actual test. The first two are
bugs in gm2 w.r.t handling a module within a procedure. This is
now fixed in CVS btw..

I'll work on bootstrap, fixing the filename issues reported
and applying your patches..

Well done again, I'm sure the Cygwin port will be of interest to many
people,

Gaius