Re: [jamboost] Command length limit

From: Pedro Ferreira (pedro.ferreira_at_[hidden])
Date: 2003-10-09 09:17:24

Volodya,

> > I'm using BBv2 on Windows XP and I've been having a few problems related
to
> > the limit on the command-line length in one of my projects, which has a
> > relatively high number of dependencies (>15):
> >
> > 1) msvc.compile.c++ started to fail because all the '-I <path>'
> > instructions caused it to exceed the limit. I got away with it by
> > explicitly changing the hardcoded limit defined in 'int maxline ()'
> > (jam_src/execnt.c). Would it be a good idea to use a mechanism similar
to
> > the one used for libs (.rsp files)?
>
> FWIW, the CVS HEAD no longer adds unreasonable number of includes paths to
> command line. This should help.

Great.

> As for RSP file --- I believe I've tried to
> pass -L option via RSP some time ago and it had absolutely no effect,
which
> led me to believe that options in response file are ignored.

Ok. This means that we are stuck to a hard limit on the number of
dependencies.

> > 2) clean started to fail for the same reason. I solved this removing
> > 'together' in 'common.jam', line 108. Is this the best approach?
>
> No, not really. This will cause each clean action to be executed
separately
> and that's not nice. I believe that "piecemeal" modifier on the same
action
> should cause bjam to feed targets in chunks which do not OS' command line
> limit.

That is what I thought.

> wonder why this did not worked. Maybe, XP has lower limit compared
> to something else?

I don't think so.

Right now I am quite busy but, as soon as I find some time, I'll try to
investigate this.

Many thanks,

Pedro