Re: [Texmacs-dev] Patches applied for next version

[David Allouche]

On Mon, Nov 25, 2002 at 01:17:34PM +0100, Joris van der Hoeven wrote:
> 
> > > 682: this should first be tested more thoroughly with
> > >      several versions of Guile.
> > 
> > I hope you will soon be doing so.
> 
> No, that is your job, since you made the patch.

I did not made the patch. A user submitted it. Actually you care much
more than I for compilation on Sun systems.

Well. That is going to stay in my tree until you like it. Now I can
carry this kind of patch over mainstream releases for very little
work.

> > Also can you explain what is the rationale behind using "generic" by
> > default instead of a more generally useful style. I do not think
> > "generic" is what can be called a "reasonable default".
> 
> I disagree; I do not see what "reasonable" objections you have.

Actually, what are the differences in available structures between
"generic" and article? If that is only the title structures then you
are right, they should be added to "generic".

If there are more differences, "generic" is not a full featured style,
so it is not generally useful, so it is not a reasonable default.

> > > 688: rejected; users want to see the complete commands being executed.
> > >      We can live with a bit of clutter.
> > 
> > I think users really do not care about the compilation messages...
> 
> The messages issued by make are expected to tell the user exactly which
> commands are being executed. I think that these commands are already
> reasonably simple in the case of TeXmacs compared to other programs.
> So I do see no reason to change this widely accepted convention.

Then I will keep it in my tree. I can compile TeXmacs dozens of time a
day, and I hate this clutter in my shell backlogs.

> > And I believe simplifying this message may cause the users to provide more
> > complete build logs when reporting problems. Did you notice a
> > difference in the quality of submitted build logs since nogencc was
> > merged?
> 
> No.

Then there is no real need to apply this patch.

> > > I also think that 595 is obsolete and that 610 was already applied before.
> > > Please remove such items, so that I do not have to look at things several
> > > times.
> > 
> > About one week ago, I added comments to these patches to explain why
> > they are still open.
> 
> Well, I will not remove the "Set text width" menu item,
> because this is an entry which is expected by many TeX/LaTeX users.
> So the update is rejected.

I strongly dislike this menu item since it is only confusing in the
absence of a full LaTeX-like margin customization facility.

Currently, you can only set top/left/right/bottom margins without
diggging in style files. Being able to additionnaly set the text-with
open the doors to different margins for odd and even pages. That is a
good feature in itself, but its semantics are completely unclear if
your are not familiar with how page margins work in TeXmacs.

So, until there is a real GUI facility to set different margins for
even and odd pages, I will keep this menu item out.

> Also, I notice that I do not like post-commenting of applied patches.
> In the case when a patch was incompletely or incorrectly applied,
> I prefer you to synchronize the patch with the newer version.
> This will greatly improve the speed with which patches are treated.

I try to do so. But sometime I forget things.

> > > I also hope that removing -I- in 683 will work for all versions
> > > of Guile and g++. This should also be tested more thoroughly.
> >
> > Testing compilation one numerous platforms is incredibly tedious and
> > time consuming. So maybe we may accept a small probability of breaking
> > a minor platform when fixing a problem on another platform.
> 
> I would call compiling TeXmacs using gcc 3.0.4 "compiling on a minor 
> platform".

Are you talking of the guile_str_size_t patch here? I do not see the
relation.

But I agree, gcc-3.0.4 is a minor platform. Yet I see no reason not to
fix it.

> > For this specific change, as far as I recall, the "-I-" flag was
> > introduced by nogencc, so not using it is unlikely to cause any
> > problem. And considering how anal the build system is about include
> > paths, I would be surprised to see many (if any) header file name
> > conflicts arise.
> 
> Well, I remember problems with the order of the include paths
> related to Guile, so I am not sure at all. As a general rule,
> I do not like "fixes" which undo previous fixes, and thereby
> introduce a big probability of error.

Unless there is a real useful changelog, it is almost impossible to
know what were the previous fixes. I would like very much to know if
that was needed for a previous fix, but currently there is no
practical way to know.

Anyway, '-I-' will prevent compilation with guile-1.6.0 so it has to
be removed. That could be done conditionnaly by the configure script,
but that is some complexity I am unwilling to introduce if it is not
clearly required.

-- 
David Allouche         | GNU TeXmacs -- Writing is a pleasure
Free software engineer |    http://www.texmacs.org
   http://ddaa.net     |    http://alqua.com/tmresources
   address@hidden  |    address@hidden
TeXmacs is NOT a LaTeX front-end and is unrelated to emacs.