Re: Is commit 2ca0d57 broken out of the box?

[Zack Weinberg]

On Thu, Feb 27, 2020 at 12:40 PM Tadeus Prastowo <address@hidden> wrote:
> Could someone try to checkout the Autoconf git repository at commit
> 2ca0d57 (https://git.savannah.gnu.org/cgit/autoconf.git/commit/?id=2ca0d57)
> and build it, please?
>
> I would like to know whether it is possible at all to build commit
> 2ca0d57 out of the box or the commit is indeed committed in a broken
> state.

This commit does indeed appear to be broken.  I get

$ make
...
autom4te_perllibdir='..'/lib AUTOM4TE_CFG='lib/autom4te.cfg'
./bin/autom4te -B '.'/lib -B '..'/lib         --language M4sh --cache
'' \
  --melt ../bin/autoconf.as -o bin/autoconf.in
Undefined subroutine &main::open_quote called at ./bin/autom4te line 261.
make[1]: *** [Makefile:2020: bin/autoconf.in] Error 1
make[1]: Leaving directory '/home/zack/projects/gnu/autoconf/build'
make: *** [Makefile:859: all] Error 2

Grepping the source tre21a3ac96e, I see many uses of a Perl function
called `open_quote` but no definition.

The fix for this appears to be in the very next commit
(21a3ac968cef962fe9dadcd7a7058f169e836e36, "Port to new
Autom4te::XFile API") which removes all of the uses of `open_quote`.
`make` works fine for me as of that commit; there are a bunch of
testsuite failures but I think they're unrelated problems.

> Please note that the commit in question makes no change to any
> ChangeLog file, and therefore, I rely on your kind help to make sense
> of the committed change whose commit log is just the terse message:
> "make fetch".

"make fetch" updates a bunch of files that are maintained as part of
other software but are also included into the autoconf source tree.
One of those files is lib/Autom4te/FileUtils.pm, which *used* to
define open_quote, but stopped doing so at some point in its own
development.  Commit 2ca0d57 brought in a version of FileUtils.pm that
no longer defined open_quote, and then 21a3ac9 fixed up the callers in
autoconf's source tree.

Re your other thread, I do not know why automake's test suite has
problems with autoconf after this point but not before.  If you are
bisecting, I would recommend you single-step forward in autoconf's
version history until its own testsuite is fully green again for you,
and then see if that also makes automake's testsuite happy.

zw