Re: [PATCH] Re: Testsuite races -- updated

[Roland McGrath <roland at hack dot frob dot com>]

> I can't see why we're creating the datafile, then moving it into place
> as separate steps in this code to begin with.  ISTM we should just have 
> msgfmt put the bits were they belong.

In general there are two reasons for this.
1. atomic supersede
   This only matters if there might be concurrent users of the old file.
   It doesn't apply here.
2. avoid leaving partial output on failure 
   This matters for anything that's a make target, since the next make run
   will see that the file exists and is new enough, and not run the
   commands again.  This applies especially to redirections, where the
   command will create (and touch) the output file even if it fails
   utterly.  I don't know whether msgfmt reliably avoids touching (or else
   removes) its output file in failure cases.  But it never hurts to cover
   this case.

> 2012-08-17  Jeff Law <law@redhat.com>
> 
> 	* intl/Makefile (codeset_mo): New variable.
> 	$(codeset_mo): New target.

Missing parens around this.

> 	(tst-codeset.out): Depend on that.  Remove explicit rule.
> 	(tst-gettext3.out, tst-gettext5.out): Likewise.
> 	(LOCPATH-ENV, tst-codeset-ENV): New variables.
> 	(tst-gettext3-ENV, tst-gettext5-ENV): Likewise.
> 	* intl/tst-codeset.sh: Remove.
> 	* intl/tst-gettext3.sh, intl/tst-gettext5.sh: Likewise.

Don't separate file names with a comma.  Use separate lines.

> +$(codeset_mo):
> +	$(make-target-directory)
> +	msgfmt -o $@ tstcodeset.po

Unless you want to examine msgfmt and testify to its inability to leave a
partial output file on failure, make this $@T with a mv -f command.

> +$(objpfx)tst-codeset.out: $(objpfx)tst-codeset $(codeset_mo)
> +$(objpfx)tst-gettext3.out: $(objpfx)tst-gettext3 $(codeset_mo)
> +$(objpfx)tst-gettext5.out: $(objpfx)tst-gettext5 $(codeset_mo)

You don't actually need the first dependency here.
The pattern rule will supply it (and make it first so $< gets it).


Thanks,
Roland