[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Sporadic davautocheck failure in trunk? (was: Re: [PATCH] Fix an inefficient way to fill an array of inherited properties)

From: Daniel Shahaf <d.s_at_daniel.shahaf.name>
Date: Mon, 25 May 2020 00:38:40 +0000

Nathan Hartman wrote on Sun, 24 May 2020 18:28 -0400:
> On Sun, May 24, 2020 at 6:00 PM Daniel Shahaf <d.s_at_daniel.shahaf.name> wrote:
> >
> > Nathan Hartman wrote on Sun, 24 May 2020 12:10 -0400:
> > > On Sat, May 23, 2020 at 8:38 PM Branko Čibej <brane_at_apache.org> wrote:
> > > > For my part, I just didn't get around to it. I think it should be committed.
> > >
> > > I committed it in r1878084.
> >
> > Thanks, Nathan. I'd have done this myself, but my build environment
> > had bitrotted to the point that davautocheck didn't pass.
>
> Uh oh, that sounds bad.
>

Well, the 1.14 pre-releases built fine for everyone else (probably
including several developers on OS's similar to mine), the bots were
green, and I had just freshly re-installed the apache2 packages, so I
figured it was _probably_ a bug in my environment.

But yeah, it's a probability.

I haven't kept fails.log, unfortunately. I built on Debian stable.
I'd have built with my usual array of flags¹, run «make davautocheck»
with «APACHE_MPM=event THREADS=1 CLEANUP=1», and confirmed that the
failure happened without the patch as well (otherwise I'd have posted
then). From memory, the failure happened approximately once per
davautocheck run (with the «TESTS» Makefile parameter unspecified,
i.e., once per full run), not always at the same place.

> I haven't looked at davautocheck.sh before. Does it test anything that
> tools/dev/unix-build/Makefile.svn doesn't?

davautocheck.sh is just a script that sets up httpd and runs the test
suite with a --url= argument pointing to that httpd.

I don't know how «make davautocheck» compares to «make -f …/unix-build/Makefile.svn
serf-check».

davautocheck.sh has some development-oriented flags, documented in
comments at the top of the script.

> Asking because I ran the latter on latest trunk with this change
> applied. (All tests passed.) Also from reading the code, including
> spelunking into the APR calls etc, I see no functional difference
> (except efficiency as stated).

"except efficiency"?

Cheers,

Daniel

¹ «CONFIG_SHELL=/bin/dash /path/to/configure AR_FLAGS=cr RUBY=none
  --disable-nls -q -C --enable-maintainer-mode 'CFLAGS=-DSVN_DEPRECATED=
  -DSVN_FS_FS_DEFAULT_MAX_FILES_PER_DIR=3 -DPACK_AFTER_EVERY_COMMIT
  -DSVN_UNALIGNED_ACCESS_IS_OK=0 -DSUFFIX_LINES_TO_KEEP=0 '», builddir
  on a tmpfs. With these flags, three svnadmin tests are _expected_ to
  fail; those failures are unrelated to the other failures discussed
  above.
Received on 2020-05-25 02:39:03 CEST

This is an archived mail posted to the Subversion Dev mailing list.