Re: GnuCOBOL 3.1-rc1 on FreeBSD, NetBSD and others

[Simon Sobisch]

Am 13.07.2020 um 18:43 schrieb Bruno Haible:
> Vince,
>
> I'm CCing bug-gnucobol because Simon has asked for feedback from testers,
> and therefore I don't want to discuss things in private with you,
> excluding him from the audience.

Thank you for CCing me and the bug list in this case.

>
> Vince Coen wrote:
>> It did pick one - -   DB is the default.
>>
>> BDB is the first one supported over 10 years ago.
>> Well over.
>
> Citing Simon's steps for testing:
>
>   2 check DEPENDENCIES for the mandatory (mostly gcc + gmp or mpir) and
>   optional dependencies (berkeley-db, cjson, libxml2, gettext)
>
> Since it's an optional dependency, I expect the configure script to
> succeed if that dependency is not present.

It is optional as in "not needed to build GnuCOBOL", and as noted in
DEPENDENCIES actually the default (so you need to explicit disable it if
you don't want it). It was my mistake to not be more clear about that,
but the good thing here: it has more awareness now and may be dropped
(like: checked if found as working, otherwise not being used) even for
the next 3.1 iteration (it was planned for 4.0 in any case).

We'll discuss this between the developers - stay tuned in the release
announcement.

>> Now I cannot find where the bug feature is at the gnu website but this
>> site is NOT monitored
>> for bug reporting or anything else for that matter by the GnuCOBOL team.
>>
>> I suggest you go to the GC website at :
>>
>>
>> www.sourceforge.net/projects/open-cobol/
>>
>>
>> There is an option top, middle to right marked as Tickets then bugs.
>
> When "make check" failed, it told me to send a mail to bug-gnucobol.
> If the GNU Cobol team doesn't want to receive these mails there, they
> should change the output of "make check" when it fails.

This list is totally monitored (just less often used than the issue
trackes, which are less used than the discussion boards).

It therefore is also correctly mentioned in the testsuite script and
cobc --help; we may should add a note to the automatically generated
testsuite, for the issue tracer option like we have in cobc (but I guess
this means hacking the generated script).

>> You only need to issue one per problem regardless of different platforms.

@Vince: maybe you can check the testsuite results and create one issue
per fail, noting the failing platforms?

>> That said this is not a bug.

Actually, it is "as it was before" and documented, too -so no bug; but
it may should be made "optional up-front" like gettext, curses and
libxml are sooner (3.1) than later (4.0).

That said: thanks @Bruno for the plattform testing and thanks @Vince for
taking care to best knowledge,

Simon