Re: [bug-gawk] "make distcheck" fails to to files left in the build directory

[Stefano Lattarini]

On Tuesday 11 October 2011, Aharon Robbins wrote:
> Hi.
>
Hello Aharon, thanks for the quick reply.

> Thanks for the mail.  Why is _beginfile2 left over?  That is the real
> question.
>
It seems to be left over in VPATH builds only.  And the reason is that
the test that creates it, i.e., `beginfile2', *fails* in a VPATH build
(but the make exit status doesn't relect this failure, yikes!)

Here is the reason of the failure, I guess:

  $ cat _beginfile2
  --Test 1a--
  gawk: fatal: cannot open file `beginfile2.in' for reading (No such file or 
directory)
  --Test 1b--
  --Test 2--
  gawk: fatal: cannot open file `beginfile2.in' for reading (No such file or 
directory)
  ...

So the problem seems simply due to a testsuite weakness (the test
`beginfile2' doens't play nice with VPATH builds).

> version.c is left around on purpose; it's needed for
> some of the non-Unix systems which don't run configure.
>
Then you should define `distcleancheck_listfiles' so that `version.c'
is explicitly ignored and don't cause a spurious failure.  You probably
haven't noticed this error before because, due to a bug in automake
1.11.1 and earlier, a single file left in the build direcotry don't
trigger a "make distcheck" failure; but this has been fixed in the
`maint' branch of the automake git repository, and the fix will appear
in the next minor automake release (1.11.2, hopefully to appear in a
month or so); so, if you start to use this new version, it will cause
the gawk "make distcheck" to fail.

> And just out of curiosity, why are you running make distcheck?  I'm the
> only one who should be doing 'make dist'. :-)
>
I was trying out the development version of automake with a few real-life
packages (gawk among them) to ensure we haven't introduced any glaring
breakage or incompatibility.  So you shouldn't consider yourself the
only one who should be doing 'make distcheck' for gawk -- count in also
the automake developers ;-)

> Thanks,
> 
> Arnold
> 

Best regards,
   Stefano