Re: warning messages and correct output

[Karl Berry]

Hi Patrice,

    Other recent examples are ... dropped now

So, is gdb's @item-less @table the only actual example at this point?
I will send a separate msg about that one at some point.

    When a warning deals with a dubious construct, but a construct that
    nevertheless could lead to correct output in some format, 

In principle, I could conceive of having some warnings that are only
given when the problematic output format is being generated, but I think
it should be a very rare case.  Texinfo is supposed to be all about
one source/multiple outputs.  None of the examples you've given seem
like obvious candidates for such treatment to me.

    There are already possibilities to turn of pointer validation with
    --novalidate

Only for historical reasons.

    Some warnings also do not lead to wrong output in any output format

Again in principle, warnings aren't just about wrong/invalid output, but
about helping users diagnose things which are always mistakes even if
technically allowed by the language.  Like writing
  if (x = 3)
in C.

Like you, as far as I can tell :), I certainly don't want to give
warnings for *useful* constructs that follow the letter and spirit of
Texinfo.  I don't think we give many warnings.  At this point, all the
warnings I have seen when working on the various documents I maintain
have been valid and useful.

    (for example @-commands that do not appear at line beginning,

I'm not sure how necessary that warning is.  TeX doesn't mind, in
general.  Such leading spaces are simply treated as a space.  Not
counting verbatim environments and the like.  Maybe we should adjust the
doc instead.

    Printindex before document beginning

I don't remember why we added that, but it seems apropros enough, since
no real document would ever want the index to come out first.  Was some
user confused about @printindex vs. @synindex or something?

    since these warnings are annoying some users, 

Did I miss something?  Are there other pending cases of actual user
complaints about real documents besides Eli/gdb/@table?  I don't recall
seeing any.

Anyway, I think we should see what happens when more users get their
hands on the code.

In general, I want to present this first pretest as "looking for
feedback" rather than "set in stone".  That is, if new warnings turn out
to be problematic, or incompatibilities show up, or whatever, then we
want to fix them in whatever the right way is.  That's why we have
pretests.

Thanks,
k