Re: Performance issue of libtool-2.4.4

[Tom Ghyselinck]

Hi everybody,

On vr, 2015-02-06 at 10:46 +0100, Peter Rosin wrote: 
> On 2015-02-06 10:30, Gary V. Vaughan wrote:
> > Hi Peter,
> > 
> >> On Feb 6, 2015, at 9:22 AM, Peter Rosin <address@hidden> wrote:
> >>
> >>> On 2015-02-04 15:48, Bob Friesenhahn wrote:
> >>>> On Wed, 4 Feb 2015, Robert Yang wrote:
> >>>>
> >>>> When reporting a bug, please describe a test case to reproduce it and
> >>>> include the following information:
> >>>>
> >>>>      host-triplet:   $host
> >>>>      shell:          $SHELL
> >>>>      compiler:       $LTCC
> >>>>      compiler flags: $LTCFLAGS
> >>>>      linker:         $LD (gnu? $with_gnu_ld)
> >>>>      version:        $progname (GNU libtool) 2.4.5
> >>>>      automake:       `($AUTOMAKE --version) 2>/dev/null |$SED 1q`
> >>>>      autoconf:       `($AUTOCONF --version) 2>/dev/null |$SED 1q`
> >>>
> >>> Perhaps libtool is accidentially executing 'automake --version' and 
> >>> 'autoconf --version' every time it is executed?  That would certainly 
> >>> lead to a huge slowdown.
> >>
> >> That's it of course, how else could the variable be assigned?
> > 
> > Only when --version is being serviced.
> 
> Are you saying the a script with this line in it:
>       foo="`($AUTOCONF --version)`"
> will delay the exec of $AUTOCONF until foo is expanded?
> 
> I think not.

No, shell variable expansion is not delayed.

Unlikey when used in `Makefile` with GNU make, where this *is* delayed
until full expansion.

> 
> >> But is it even useful information? I would expect that the autofoo
> >> versions *at the time the package was created* is what matters?
> > 
> > The information is useful in bug reports, and our instructions for 
> > reporting a bug to the list explicitly ask for the output from `libtool 
> > --version` which by including their other autotool versions makes 
> > reproducing the reporters environment a lot easier :-)
> 
> But are the autofoo versions at libtool time really what we want
> to know in bug reports? Again, I'd be much more interested in the
> autofoo versions used to bootstrap the package. That might often
> be the same thing, but when they are not confusion will ensue.

AFAIK: Doesn't automake --version depend on the project's configure.ac?
At least some distribution have an "automake-wrapper" which selects
the automake version based on the project settings and fall-back to
system default.

e.g. Running "automake" (i.e. "automake-wrapper") when you have
"automake-1.10", "automake-1.11" and "automake-1.12" installed:

- Suppose system-default is automake-1.11

- When executed in directory *without* `configure.ac`: automake will run
automake-1.11.

- When executed in directory *with* configure.ac:
  - When automake version is *defined* to '1.10' in `configure.ac`,
    it will run "automake-1.10"
  - When automake version is *defined* to '1.12' in `configure.ac`,
    it will run "automake-1.12"
  - When automake version is *not defined* in `configure.ac`,
    it will run "automake-1.11"

With best regards,
Tom.


> 
> Cheers,
> Peter
> 
> 
> _______________________________________________
> https://lists.gnu.org/mailman/listinfo/libtool

-- 

________________________________________________________________________


| address@hidden
|
| Tom Ghyselinck
| Senior Engineer
| Excentis N.V.
| Gildestraat 8 B-9000 Ghent, Belgium
| Tel: +32 9 269 22 91 - Fax: +32 9 329 31 74

________________________________________________________________________