Karim Belabas on Sat, 29 Nov 1997 21:34:36 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: Os/2 portablity patch for pari2.0-alpha


Hi Ilya,

   while preparing the 2.0.1.alpha update, I finally had a look at your OS/2
patch (msg 19).

1) I think your diff was not complete: it can't possibly work the way it is
now. Presumably, modifications to Configure are missing (we need to set
osname=os2 somewhere. As well, your introduce variables make_sh, shell_q,
ln_s, exe_suff which are not defined anywhere).

2) your change of doc/Makefile to make_doc.SH is problematic. We don't want
to extract this file, since I want it to be usable even if Configure fails
completely. That's the reason why there's the ugly target "force" which
substitutes @version@ (we don't want to expand this particular @version@).

Some notes:

> a) gcc is prefered to g++, is it correct?

It is. There should not be any speed difference between the two now (did some
benches when I introduced the _inline_ hack, last year g++ produced slightly
slower code on at least two different machines. Of course, now somebody else
is going to redo it and prove me wrong...)

> c) Version check for readline was checking for older variable
>   LibraryVersion, not present in newer readlines;

  This was only used to address a bug in readline 2.0 on Solaris (SIGINT
blocked by readline and not properly restored). This bug has been corrected
in version 2.1. So we DO want to check for LibraryVersion... But since GP
does not use anymore any features of readline 2.0, I'm just cancelling this
completely. We don't check at all for the version number (and do a possibly
unnecessary sigrelse/sigsetmask in gp.c).

Cheers, Karim.
--
Karim Belabas                          e-mail:
Max-Planck-Institut fuer Mathematik       karim@mpim-bonn.mpg.de
Gottfried-Claren-Str. 26               tel:
53225 Bonn (Germany)                      (00 49 228) 402-245