Re: Problem with new "make bootstrap"

[Andris Pavenis <pavenis at latnet dot lv>]

On Sat, 16 Sep 2000, Richard Kenner wrote:
> Once it finishes, it never realizes that files have changed.  So it's
> not like the old "make bootstrap3" where it would do any needed recompiles.
> You now have to remove stage_e, which is very counterintuitive.  Unless
> some more automation can be added to allow the normal kinds of things we
> need to do to be done without removing mysteriously-named files, I'd
> suggest this change be reverted: the confusion isn't worth the improvement.

As far as I remember one of ideas behind restartable bootstrap was DPMI 
problems when building DJGPP port of gcc under Win9X (LFN support is 
currently required). 

With gcc-2.95.X I choosed different way - splitting bootstrapping in separate 
smaller parts and running them from DOS bat file. It may look ugly but it 
worked Ok under Win9X. So I was able to drop original
source archives (gcc-2.95.X.tar.gz and libg++-2.8.1.3) into directory 
where my scripts and bat files were located, start one bat file and after
some time I got distribution archives for DJGPP (without any my 
intervention in build time unless something bad happened)

My experience showed that restarting make (I'm talking about DJGPP) is not 
always successfull. For example if I got crash due to running out of 
descriptors while configuring one of libraries (eg. libstdc++) and Makefile
were created but not written up to end then restarting make failed. (I 
haven't verified this for a rather long time tough as the way I used for 
gcc-2.95.2 was different)

Andris