On Wed, Dec 21, 2011 at 09:29:31AM -0000, Phil Holmes wrote:
I'll hold on this one and look again once Julien's patch is pushed.
As I said, it's almost inconceivable that there would be an error
from texi2pdf causing the build to fail.
I've seen half a dozen in the past month. (my fault, it was
problems in @example vs. @verbatim).
As a default, it tolerates 1000 errors before terminating. Most
of our translated docs already have a few hundred errors which
would cause the build to fail if we set the --error-limit to a
lower value.
Are you sure? I'd expect it to tolerate 1000 *warnings* before
terminating. Warnings are not errors.
...
hmm, as far as I can see, texi2pdf doesn't have any --error-limit
flag, but makeinfo does.
$ man makeinfo
...
General options:
--error-limit=NUM
quit after NUM errors (default 100)
WTF?! David: is this seriously giving a zero return value if
there's any "errors" ? Or does it still return a zero, but print
out multiple errors?
Part of the difficulty in testing this patch was trying to get
texi2pdf to fail so I could see what happened!
Easy way to make it fail immediately: add
@example
@\
@end example
to some file. I recommend something in the CG.
makeinfo will not fail, but texi2pdf will fail.
What do you see if you set --error-limit=0 for makeinfo? I'm
wondering if this is why we sometimes see "make" compiling the
docs but "make doc" failing. If we can get "make" to be stricter
about checking the texinfo, this could save a lot of time in
writing docs.
If it would take less than 10 hours to fix the english docs, I
think we should set --error-limit=0. We can leave it at 100 for
the translations for the next X months.