[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gnuplot-info-beta
Subject:    Re: Three small patches
From:       Hans-Bernhard Broeker <broeker () physik ! rwth-aachen ! de>
Date:       1999-02-23 19:15:16
[Download RAW message or body]

On Tue, 23 Feb 1999, Lars Hecking wrote:

> > 3) The 'sleep' function linked in by makefile.mgw is an incorrect one: 
[...]
>  Just to make this clear: undefining HAVE_SLEEP causes the sleep() function
>  from stdfn.c to be linked into the executable, which in the case of WIN32
>  multiplies it's argument by 1000 before passing it on to Sleep().

Yes.

>  From this I gather that MinGW's sleep() is just an interface to Sleep()
>  on Win32, correct? 

Yes. It's an old name of the same function, indeed. In particular, it does
not multiply its argument by 1000, as it should, and it has a slightly
different prototype, leading to collisions when gnuplot shoots out its own
prototype. That's what the new -D_NO_OLDNAMES is needed for: it masks out
the old name 'sleep' for 'Sleep()'.

To me, this seems to be a bug in the MinGW32 headers: if any function is
in there, with a name reserved to any ANSI/POSIX function, it shouldn't
ever have a different type than the standardized one, at least not
with default compiler switches.

> If this is the case, I'd rather keep the HAVE_SLEEP
>  define in the makefiles, and add another branch to the GP_SLEEP definition
>  in stdfn.h. Would this work?

I doubt that (unless you're willing to 

#define GP_SLEEP(seconds) Sleep(1000*seconds)

OTOH, we already have the necessary branching machinery inside stdfn.c's
definition of 'sleep', for all WIN32 compilers. We just have to use it, by
not #defining HAVE_SLEEP. 

>  I have not heard from Dave Kotz yet about possibly moving from
>  ftp.dartmouth.edu to ftp.cs.dartmouth.edu.
> 
>  Would we need to get TW's ok for a bugfix release?

Depends on which clause we want to publish it under. If
we don't ask TW, then it's not an 'official' release, and
we'ld be bound by the 'modified version' terms:

'0.' says we can't distribute a gnuplot-3.7.patch1.tar.gz source
package :-(

1. says we'ld have to include that 500K+ source patch with each
of the binary packages (all 10 or so of them!), probably scaring
some users out of their wits :-(

2. says we must add a line like 'gnuplot-beta-team special version
patchlevel 1' to the startup message.

3. says we'ld have to change the 'primary contact' address to
point to either info-gnuplot-beta, or Lars' own email.

The longer I look at this list, the less I like it...

Hans-Bernhard Broeker (broeker@physik.rwth-aachen.de)
Even if all the snow were burnt, ashes would remain.


[[[[ unsubscribe from info-gnuplot-beta via majordomo@dartmouth.edu ]]]]

[prev in list] [next in list] [prev in thread] [next in thread]