Re: [GnuCOBOL 3.1.2] testsuite: 6 failed

[Simon Sobisch]

Am 23.10.2021 um 00:18 schrieb Jesus Vargas:
> Hi, I just came to report some bugs (not too much, just those ones).

One may could consider this one minor bug in the testsuite.

> I have to recognize that this is one of the few projects that have
> successfully compiled under Tiny C compiler. Thanks for coding in this
> clean and standardized way, and keep going!

Thank you for the kind words. A high level of C portability is one of
the goals of GnuCOBOL. It could be possible that TinyC may not be
"supported" forever, but at least with 3.x it definitely will (although
it isn't tested by anyone regularly, so it would be good to subscribe to
the user list to get the announcements of an upcoming version and
recheck then or from time to time with the nightly builds [1]).

> As additional information, my Tiny C package was installed from the AUR
> repos and compiled with clang. Even though I could use clang for
> compiling OpenCobol but I was testing, as well as I knew (by experience)
> that this is one of the best coded software because it can be compiled
> also under android using just termux and clang (also TCC compiles under
> android but with some bugs). #BTW I use Arch... moreless, I'm using
> Manjaro Linux in its latest version.

Do you have a recent build on Android or can test that, please? My last
Android build of GnuCOBOL was years ago. (I assume you had tried with
GnuCOBOL as you did this time on your TinyC environment).

> Even if it is needed, I have the files generated in with "make test"
> command, but I'm not sure which of them to send.

testsuite.log has all the configuration details and the one unexpected
failure (there are "expected" ones as honest note where we know GnuCOBOL
does not do the correct thing currently [these commonly are no new bugs,
but something that was never implemented fully "yet"; not a reason to
not use GnuCOBOL]).
The output of make test you've sent also shows that everything is fine
there:

Comparing total test results...
diff summary.txt...
Done

--> so no unexpected failing test

> Anyway I'll install as it is, as well as I don't require compile assembler 
> files.

Actually the part of the test for compiling assembler would have worked
as far as I know.
TinyC has no option to generate assembler files so that test is kind of
expected to fail. I've did some changes after your mail, so `cobc -S`
should now error on "normal" runs and send the testsuite a "skip" return
code. You may want to try it (I haven't verified that it works soa
quick test would be appreciated) with the nightly build, deep link for a
.gz tarball at [2] (you can replace gz with xz or lz or bz2 if you want
one of those formats)


> Greetings from Mexico.

Greetings from Germany back!

It would be nice if you could report if 3.2dev has no unexpected failure
on your environment now, and possibly how an Android build of it works
and if that passes  make test (if you have perl available there) + make
check.

Simon


[1]
https://sourceforge.net/projects/gnucobol/files/gnucobol/nightly_snapshots/
[2]
https://ci.appveyor.com/api/projects/GitMensch/gnucobol-3-x/artifacts/gnucobol-3.2-dev.tar.gz?job=Image:%20Ubuntu2004