[Swan-dev] build dependency for version.c is broken ccd2cf87 ?

[Andrew Cagney]

Yes and no - there are two things at play:

- the obvious problem is that the generated file version.c shouldn't
even be in the source tree

While the commits:

commit dc49a97017876d89d9cb144298c971a7f74f1073
Date:   Thu Apr 13 15:23:28 2017 -0400

    building: switch pluto to using mk/program.mk

    Still has hack to include mk/config.mk

    Old pluto Makefile's clean target didn't delete all .o files
    leading to some strange behaviour.

commit ccd2cf870ea0628e3689321a8f5564487577193b
Date:   Thu Apr 13 14:31:38 2017 -0400

    build: switch the libswan.a build to using mk/library.mk

    Fixes a dependency regression with ike_alg_serpent.c et.al.

    Still has a config.mk hack.

fixed the reason behind the recent ike_alg_ripemd.o missing breakage -
"make clean" in programs/pluto/ and lib/libswan/ didn't do a proper
job and and could leave old ike_alg_ripemd.o files behind (see also
00de09c5f3af73b2a1ccbc9708834b81348c14a4 and
85ec9e62498ac1f32e741c5a002e55801756bf14) - they inadvertently
resulted in the generated file "version.c" ending up in the source
tree (I'll blame VPATH :-^).  Restoring the files location and
ensuring it is deleted will restore the old behaviour.

- version.c, regardless of location, does not have any dependencies

This was discussed here
https://lists.libreswan.org/pipermail/swan-dev/2016-August/001609.html

On 25 April 2017 at 01:54, Antony Antony <antony at phenome.org> wrote:
> I noticed libreswan build system is keeping a stale copy of lib/libswan/version.c possibly since commit ccd2cf. Also another related one is  modobj/version.c. The second one is probably due to a different issue. And it probably has a longer history.
>
> To demonstrate the issue I picked a test case randomly.
> The version reported in road.pluto.log and test directory are not the same.
>
> in road.pluto.log v3.20-159-gccd2cf8-master
> test directory suggest 3.20-178-g9ec5d42-master
>
> http://testing.libreswan.org/results/testing/3.20-178-g9ec5d42-master/xauth-pluto-19/OUTPUT/road.pluto.log.bz2
>
> It has been going on for over a week, on my testruns I would think something wrong with compile because pluto is recording old version. Start to blame 9p etc.... And re-run again and again.
>
> Now my guess ,as the this log suggest,  is lib/libswan/version.c's dependency is broken with commit ccd2cf.
>
> -antony
> _______________________________________________
> Swan-dev mailing list
> Swan-dev at lists.libreswan.org
> https://lists.libreswan.org/mailman/listinfo/swan-dev