Re: Second patch ping

[Zack Weinberg <zack at codesourcery dot com>]

Ian Lance Taylor <ian@wasabisystems.com> writes:

> You suggested that pex_init would take an argument of PEX_USE_PIPES.
> I didn't implement that, and the code always uses pipes on Unix.  gcc
> already has temporary file handling, and it needs control over it in
> order to implement --save-temps.  I didn't see any special advantage
> to having the file handling in both pexecute and in gcc.c.

Hoom, see, I was hoping to get the file handling out of gcc.c (by
relying on all the tools' ability to read/write standard
input/output).  I had forgotten about save-temps mode though.

*scratches head* To get this right we would need to be able to specify
the file extension for the target temporary file on a per-pex_run()
basis, and "use this basename" or "pick a random temp basename and
delete when done" modes on a per-pex_init() basis.  It may well not be
worth it, but argh, I like the idea of getting a huge pile of gunk out
of all the compiler specs.  Thoughts?  I don't want to have to change
pex_run() later (or invent a new entry point).

zw