Re: Incorrect checking of '-g'

[Bram Stolk]

On Mon, 31 Dec 2001 14:32:52 +0100
Bram Stolk <address@hidden> wrote:

OUCH!
I made a mistake. -c -o is not used for -g check.

However, the -g check *does* fail for the icc compiler.
But this is because icc prints the source filenames on stdout
while compiling.

the -g check assumes that all output from the compiler
will indicate problems, and will therefore conclude that
-g is illegal.

Is an ANSI C compiler allowed to produce output on stdout
or stderr, even if the compile is 100% ok?
If so, the -g check, as is, is not conclusive.

Sorry for drawing the wrong conclusions too fast.
However, I do think that libtool should be compatible with icc.
I could not find a cmdline flag to make icc silent about
the input files it is processing.
Maybe the -g check should be improved afterall?

thanks,

  Bram
 

> Hello,
> 
> 
> There is a severe bug in libtool version 1.4.2
> libtool checks for both C and CXX compilers whether
> the '-c -o' contruct is allowed.
> 
> Even if it finds that this is not the case,
> it will later on check for '-g' capability USING THE VERY SAME
> '-c -o' construct.
> 
> Ofcourse, the test will fail (due to '-c -o', not because of '-g')
> 
> This results in an invalid assumption, causing GNU autoconf/automake
> enviroments to skip debug flags when using e.g. the new Intel icc 
> compiler for linux.
> 
> affected file:      /usr/share/aclocal/libtool.m4
> affected version:   1.4.2
> 
> 
>    Bram Stolk