Re: [jamboost] updated mipspro-tools.jam, etc.

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-06-06 13:18:52

----- Original Message -----
From: "Ralf W. Grosse-Kunstleve" <RWGK_at_[hidden]>
To: <jamboost_at_[hidden]>
Cc: <rwgk_at_[hidden]>
Sent: Thursday, June 06, 2002 2:08 PM
Subject: Re: [jamboost] updated mipspro-tools.jam, etc.

> The last couple of days I have been trying to compile Boost.Python
> using mipspro-tools.jam. Eventually I had to rewrite most of the file.
> The current result is attached. This works for almost all of
> Boost.Python (I believe the remaining problems are not specific
> to mipspro-tools.jam).
> The file below does away with the hard-wired system paths.

What do you mean by "hard-wired system paths"?

> IMO it
> is much better to rely on the compiler for providing these.
> It should be possible to use another ABI with e.g.
> -sMIPS_ABI="-n64 mips4", but I haven't actually tried it.

I think you're missing the use of features here, but I'll let Dan Gohman
explain.

> Is there someone who feels responsible for maintaining
> mipspro-tools.jam in the central boost cvs tree? If not, I could
> assume this role. If there are suggestions for improving the
> attached file, please let me know.

My only remarks:

1. CC_CMD is too-simple a global variable name for use in an environment
where people may compile with multiple toolsets with a single command.
2. It should be possible to configure the path to the compiler installation
so that users are not required to have the compiler in the path before
building. When builds use multiple toolsets it can be way too capricious to
try to prepare the environment for all of them simultaneously.