Re: Fixing testsuite failures: arithX.mod

[Gaius Mulley]

Brian Callahan <bcallah@protonmail.com> writes:

>> Brian Callahan bcallah@protonmail.com writes:
>>
>> > Hello --
>> >
>> > There are many testsuite failures on OpenBSD in the arithX.mod
>> >
>> > compilation. They all boil down to the same issue: CMPLX is being
>> >
>> > used in a CONST assignment but gm2 does not believe that is valid.
>> >
>> > Attached is the full error message for arith3.mod. If there are
>> >
>> > any pointers, I can go fix it.
>> >
>> > Thanks.
>> >
>> > ~Brian
>>
>> Hi Brian,
>>
>> I'm curious as to the contents of <builddir>/gcc/testsuite/gm2/gm2.log
>>
>> regarding arith3.mod. On my debian buster amd64 system I have the following:
>>
>> spawn -ignore SIGHUP
>>
>> /home/gaius/GM2/graft-combine/build-gm2-master-enabled/gcc/gm2
>>
>> -B/home/gaius/GM2/graft-combine/build-gm2-master-enabled/gcc
>>
>> -I/home/gaius/GM2/graft-combine/build-gm2-master-enabled/x86_64-pc-linux-gnu/./libgm2/libm2iso:/home/gaius/GM2/graft-combine/gm2-floppsie/gcc/testsuite/../m2/gm2-libs-iso
>>
>> -I/home/gaius/GM2/graft-combine/build-gm2-master-enabled/x86_64-pc-linux-gnu/./libgm2/libm2cor:/home/gaius/GM2/graft-combine/gm2-floppsie/gcc/testsuite/../m2/gm2-libs-coroutines
>>
>> -I/home/gaius/GM2/graft-combine/build-gm2-master-enabled/x86_64-pc-linux-gnu/./libgm2/libm2pim:/home/gaius/GM2/graft-combine/gm2-floppsie/gcc/testsuite/../m2/gm2-libs
>>
>> -fiso
>>
>> -L/home/gaius/GM2/graft-combine/build-gm2-master-enabled/x86_64-pc-linux-gnu/./libgm2/libm2iso/.libs
>>
>> -L/home/gaius/GM2/graft-combine/build-gm2-master-enabled/x86_64-pc-linux-gnu/./libgm2/libm2cor/.libs
>>
>> -L/home/gaius/GM2/graft-combine/build-gm2-master-enabled/x86_64-pc-linux-gnu/./libgm2/libm2pim/.libs
>>
>> -fno-diagnostics-show-caret -fno-diagnostics-show-line-numbers
>>
>> -fdiagnostics-color=never -g -c -o
>>
>> /home/gaius/GM2/graft-combine/build-gm2-master-enabled/gcc/testsuite/gm2/arith3.o
>>
>> /home/gaius/GM2/graft-combine/gm2-floppsie/gcc/testsuite/gm2/complex/pass/arith3.mod
>>
>> the item of particular interest is the -fiso switch - which should be
>>
>> present and will enable CMPLX to be utilised in a constant expression.
>>
>> Without -fiso I get the same error - and the testsuite should be running
>>
>> these tests with the boilerplate iso configuration.
>>
>> In summary:
>>
>> gm2 -g -c arith3.mod
>>
>> will gives me the same error whereas
>>
>> gm2 -g -c -fiso arith3.mod
>>
>> is successful. An interesting problem!
>>
>> regards,
>>
>> Gaius
>
> Uploaded the whole gm2.log (gzip'd) here:
> https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fibara%2Fmisc%2Fblob%2Fmain%2Fgnu-modula-2%2Fgm2.log.gz&amp;data=04%7C01%7Cgaius.mulley%40southwales.ac.uk%7Cd510e68b218547d3d2cb08d9109fdd17%7Ce5aafe7c971b4ab7b039141ad36acec0%7C0%7C0%7C637559101057298018%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=3eg0uscBZxowXgckk9Aw9WQa5sXnxqfAOR8hxI8W0wM%3D&amp;reserved=0
>
> I believe I see -fiso in my invocation, but as I'm not too
> familiar with dejagnu, better to give you more info than less :)
>
> ~Brian

Hi Brian,

many thanks!  I can see two problems currently:

  *  it looks as if there is an unresolved shared library
     reference:  significandl which is causing a number of failures.
  *  arithX fail when running.

it would be interesting to observe the behaviour of arith3 running under
gdb - just single stepping the main program module and seeing which
assert fails - or by introducing printfs (probably better for dejagnu
log purposes).  I'll modify the testsuite (to add diagnostics) and push
the gm2 repro to savannah


regards,
Gaius