Re: [GnuCOBOL 3.0-rc1] testsuite: 17 (compile from stdin) failed (on cygwin64)

[Simon Sobisch]

Am 11.05.2020 um 02:12 schrieb address@hidden:
> Simon,
> You must be looking at the wrong file. This is the test results from the file 
> I sent:
> [...]> The snippet you pasted identified it as the testsuite.log for a
3.1-Dev test suite that ran 1012 tests instead of 895 for 3.0. I sent in
a file for the 3.0-rc1 test suite.

I was looking at and quoting the test results from a build machine (the
CI server) that auto-verifies the *current* commits on "win32 posix" space.

The error you've quoted is not shown there (but as I said other ones
are, considered false-positives for now).

> If you scroll down to line 1170 in the testsuite.log I had sent in it shows:
>
> [...]
> @@ -0,0 +1 @@
> +/cygdrive/c/Gnu_COBOL/gnucobol-3.0-rc1_win/gnucobol-3.0-rc1/tests/testsuite.dir/at-groups/17/test-source:
>  line 37: ./a.out: Permission denied
> ./used_binaries.at:552: exit code was 126, expected 0
> 17. used_binaries.at:539: 17. Compile from stdin (used_binaries.at:539): 
> FAILED (used_binaries.at:552)
>
> If I read that right it indicates there is a permissions problem with the 
> file ./a.out which is in the /test/testsuite.dir/013 folder. I would need to 
> know what permissions are required for which user to troubleshoot the problem.

It "reads" permission denied, but there are multiple reasons for it, on
win32 the one that is normally the case when running the testsuite is
that there's another process blocking the process as outlined before.

Note: you always have file permissions in this case as the complete
folders were created by running the testsuite = all by one user.

Simon



>
> Patrick Schmid
>
>
> -----Original Message-----
> From: Simon Sobisch <address@hidden>
> Sent: Sunday, May 10, 2020 4:48 AM
> To: address@hidden; address@hidden
> Subject: Re: [GnuCOBOL 3.0-rc1] testsuite: 17 (compile from stdin) failed (on 
> cygwin64)
>
> Am 10.05.2020 um 01:52 schrieb address@hidden:
>> Simon,
>> Thank you for the detailed response.
>
> You're welcome.
>
>> I have been playing with the gnucobol install for version 2.2, 3.0, and 3.1.
>> When I get to the "make check" step they all fail on the compile from
>> stdin test.
>
> I've checked the CI result for 3.1-dev with cygwin[1] which says:
>
> ------------------------------
>
> ## ------------- ##
> ## Test results. ##
> ## ------------- ##
>
> ERROR: 1012 tests were run,
> 29 failed (28 expected failures).
> 6 tests were skipped.
>
> ## ------------------------ ##
> ## Summary of the failures. ##
> ## ------------------------ ##
> Failed tests:
> GnuCOBOL 3.1-dev test suite: GnuCOBOL Tests test groups:
>
>  NUM: FILE-NAME:LINE     TEST-GROUP-NAME
>       KEYWORDS
>
>  730: run_file.at:5453   INDEXED file with LOCK AUTOMATIC (2)
>       runfile
> ------------------------------
>
> So your failing test is not included and the one failing there is a 
> false-positive (the process started from cygwin does only see it was started 
> from Win32 and therefore uses ".\prog2" for the sub-program, but then cygwin 
> translates it to ".prog2" so the startup doesn't work).
>
>> In looking at the testsuite.log each version is failing on the
>> "$COBCRUN_DIRECT ./a.out" command and it is failing with permission
>> denied on ./a.out. That is the only test that fails. As all the other
>> tests pass I must have done a successful install. That has me thinking
>> the problem could be with the configuration. What would be
>> misconfigured on my system to cause a permission denied error on the command 
>> "$COBCRUN_DIRECT ./a.out"?
> The big issue is likely the OS... I guess you may trip on another process 
> locking the executable (common "bad guys" are Windows User Experience / 
> Compatibility checker, Indexing Service or Security Suites).
> You may want to (likely temporary) disable them during the build and tests 
> and/or add exceptions to the directories and executables from your Cygwin 
> environment.
>
>> I am running on an old ASUS, their first portable that got Republic of
>> gaming certification (bought for the power and not the games) that is
>> running a currently updated Windows 10. I have installed Cygwin
>> version 3.1.4. During installation when I got to the screen to select
>> the packages to be installed I took the option to install everything.
>
> That's a huge install - I hope you did not do so only for getting GnuCOBOL to 
> work as MinGW definitely would be much smaller and reasonable for this (see 
> notes in the previous mail below).
> Of course you could - which will get you a much faster an secure environment 
> in any case - use a GNU/Linux system (dual-boot installations work quite fine 
> these days, but the setup or "general cygwin issues" are not something that 
> really belongs to this bug tracker ;-).
>
> Still thank you for the note as people now have a chance to find it.
>
>> Patrick Schmid
>
> If you have any more questions that are not related to this bug I suggest to 
> use either the user's list [2] for any GnuCOBOL related issues or discussion 
> boards [3] for this.
>
> Simon
>
> [1]:
> https://ci.appveyor.com/project/GitMensch/gnucobol-3-x-win32-posix/build/job/mua8i5552p2ekab3/artifacts
> [2]: https://lists.gnu.org/mailman/listinfo/gnucobol-users
> [3]: https://sourceforge.net/p/open-cobol/discussion/help/
>
>
>>
>> -----Original Message-----
>> From: Simon Sobisch <address@hidden>
>> Sent: Friday, May 8, 2020 12:44 PM
>> To: address@hidden; address@hidden
>> Subject: Re: [GnuCOBOL 3.0-rc1] testsuite: 17 failed
>>
>> Hi Pat,
>>
>> this error seems to have not come up here before (at least I do not
>> remember
>> it) but there are two things to know:
>>
>> 3.0rc1 is quite outdated (there was and will be no 3.0 final release).
>> The current release plan for 3.1rc1 is May/June 2020.
>>
>> I suggest to use a newer version, you can find the nightly builds'
>> address (as those are not unlikely to change in the future) listed at
>> https://sourceforge.net/projects/open-cobol/files/gnucobol/nightly_sna
>> pshots
>> /
>>   (the build history shows that the test is running now, also under
>> Cygwin)
>>
>> If you just wanted to use GnuCOBOL on Win32 (and not to extend your
>> existing Cygwin environment with GnuCOBOL) I highly suggest to just
>> use the MinGW based ready-to-use binary packages.
>> You find those at https://www.arnoldtrembley.com/GnuCOBOL.htm (a quite
>> recent 3.1-dev is available there).
>>
>> Thanks for reporting, letting others to know to possibly prefer a
>> newer and/or different environment than 3.0rc1 via Cygwin.
>>
>> Simon
>>
>>
>> Am 06.05.2020 um 23:52 schrieb address@hidden:
>>> Results from install GNU COBOL Compiler, version 3.0 RC1. Had one
>>> error, failed the compile from stdin test.
>>>
>>> Pat Schmid
>>> Chief Technical Officer, Teton Editorial Services
>>>
>>
>>
>
>