Re: Notification message levels are too high

[Ralf Wildenhues]

Hi Bruce,

* Bruce Korb wrote on Thu, Dec 01, 2005 at 12:34:32AM CET:
> On Wednesday 30 November 2005 03:22 pm, Bruce Korb wrote:
> > 
> > > libtool: install: warning: remember to run \
> > > `libtool --finish /home/bkorb/ag/ag/autogen-5.8pre7/_inst/lib'
> > 
> > Warnings are generally for the purpose of alerting people to the fact
> > that something correctable leaves the opportunity for errors or
> > hidden problems.  The string, "warning" draws peoples attention to
> > it.  In libtool's case, every warning is unavoidable and simply
> 
> Just to preempt any overly careful readers by an underly careful
> writer:  ``every warning listed below'' :-)

Do you do DESTDIR installs by default, or what's the reason you get
these warnings regularly?

> > notifications about one thing or another.  And, of course, this
> > is all the result of automated build rules that few have any
> > control over anyway.  So, please consider using "notice" instead
> > of "warning".  Thank you!  :)

I'd take a patch for CVS HEAD implementing
  func_notice
for those, and I'd also take a patch for CVS HEAD implementing
multiple `--silent' arguments, where one would turn off normal output,
and two would turn off everything but fatal errors.

The only nontrivial bit about such a patch would be that we may be
inconsistent in our use of func_error vs func_fatal_error at the moment.

Except I'm not totally sure whether
>> libtool: install: warning: remember to run `libtool --finish 
>> /home/bkorb/ag/ag/autogen-5.8pre7/_inst/lib'

would qualify as a notice only.  Just because leaving it out won't hurt
on some systems -- in general it's quite necessary.


Upon further thinking, I'd probably even take a patch for CVS HEAD that
makes libtool honor several environment variables, for the different
modes.  Something like
  LT_FLAGS
  LT_LINK_FLAGS
  LT_COMPILE_FLAGS
  ...

and then you could do LT_FLAGS='--silent --silent'.  I'm just not quite
sure about the naming yet.  It is imperative that it does not conflict
with prior use, e.g. in Automake or other build systems.

And yes, each of these features should be in a separate patch, and
including documentation -- send to libtool-patches.  The documentation
should include a notice that bug reports about libtool should generally
not provide the --silent output (unless of course the bug concerns that
option specifically).

Bruce, I don't expect you to work on any of this, but I know there has
been desire for this by others as well, mostly in this combination, so
I thought I'd mention it -- there may be someone out there whose itch is
strong enough.  :)

Cheers,
Ralf