Re: [patch] list of input files if max_cmd_len exceeded

[Ralf Wildenhues]

* Peter O'Gorman wrote on Thu, Apr 14, 2005 at 02:38:49PM CEST:
> Ralf Wildenhues wrote:
> >
> >Also, if we find more linkers with more link script syntax forms, it
> >might be good to fold that into a recipe to be put in libtool.m4.
> >Something like
> 
> I don't think that this is necessary yet. David's patch should, for 
> example, work with non-ancient darwin tools, just need to use -filelist 
> <filename>.

I don't understand why my approach should be less general or not work
with some systems.  Adding support for -filelist on darwin would be
trivial as well.

I just don't like the file_list_spec interface.  Seems too specific to
me (for the exact reason that it can't handle the GNU ld case), while
the creation of a link script seems more natural to me.  Granted, my
using $save_libobjs was stupid, but other than that?

> >Unnecessary.  Just do
> >  $ECHO "X$save_libobjs" | $Xsed | $SP2NL > $output
> >instead of looping over every file.  Note that $ECHO may only take one
> >argument.  Also, your patch does not remove $output before using, so a
> >second link would have a wrong list.
> 
> This would seem to be true for the gnu linker script case too. I think the 
> time saved by doing this is microscopic compared to the time saved by 
> avoiding the ld -r loops though :)

Hey, it's faster, less code, and IMVHO more readable, too, so why not
use it?

Regards,
Ralf