Re: updated windows installer for PARI

[Bill Allombert <Bill.Allombert@math.u-bordeaux1.fr>]

On Mon, May 06, 2013 at 11:49:47PM +0200, Karim Belabas wrote:
> * Bill Allombert [2013-05-06 21:01]:
> > On Thu, Apr 18, 2013 at 12:12:24AM -0400, jacques G wrote:
> > > > I tested on some windows server 2008 and CTRL-C worked fine.
> > > 
> > > 1. On my VISTA laptop, CTRL-C just rings the bell, while other
> > > readline controls seem to work fine. A cheap solution is 
> > > decreasing the stack size, so a runaway computation
> > > (such as an impossible task) will stop when memory is low, sooner. 
> > 
> > Further experiment by Jacques shows that disabling readline
> > (with e.g. default(readline,0)) fix this issue and the one with 
> > the cursor.
> > 
> > So I plan to add to gprc.txt 
> > readline=0.
> 
> This would be unfortunate.

Yes, but at least this does not preclude the user to activate readline if it
works on its system.

> Maybe we could try to understand
> what exactly goes wrong with MinGW's readline (or termcap ?).

Seemingly incompatibility with Vista terminal.

> - preliminary testing: does e.g. MSYS bash work out-of-the box ?
> 
> - how exactly did you get libreadline.a for MinGW ? 

I cross-compiled it together with libtermcap.

> - we could try linking with libncurses instead of libtermcap
> [ a while ago, the cygwin readline port used ncurses instead of termcap;
> the latter was broken ]

ncurses does not work on mingw.

> - we could try a few different versions of readline

This is problematic: I could not get older version to build with Mingw.

> - if nothing works and readline + MinGW remain broken maybe somebody
>   knows a *good* simple terminal that we could include in the installer

The major issue is that it works fine under wine and under MS server 2008,
which is the only version I can easily try. So basically I cannot reproduce
the issue. On the other hand there are various report of similar issues
with readline and windows on the net. 

Cheers,
Bill.