Re: [AUCTeX-devel] Re: multi-file support

[David Kastrup]

Ralf Angeli <address@hidden> writes:

> * David Kastrup (2005-05-14) writes:
>
>> Ralf Angeli <address@hidden> writes:
>>
>>> That's too drastic.  My suggestion would be to include the file
>>> name in the prompt for the master file in case the buffer is not
>>> visible.
>>
>> I disagree, actually, as you can see by the code I placed in
>> tex-buf.el for error handling.
>
> Which code?  The one in
> <URL:http://savannah.gnu.org/cgi-bin/viewcvs/auctex/auctex/tex-buf.el.diff?r1=1.225&r2=1.226>?

Yes, in chunks 1408 and 1478.

> It would be nice if you were more specific.  This way you could
> spare people some of their little time.

Sorry, I'll try to at least pick out the ChangeLog entry next time.

>> When a file gets loaded as part of processing a run on a known
>> master file, its TeX-master variable should _silently_ get set to
>> the known value (no local variable section hacking done).
>
> How do you suppress processing the local variables specifications?

I don't.  I just set TeX-master afterwards, overriding the local
variable action.

>> If the session already has different means to provide consistent
>> information for the currently loaded files, we should make use of
>> them.
>
> In case of the question for shared files I need a way to distinguish
> the file being opened by hand and programmatically in course of
> document processing.  Functions opening files during processing
> could temporarily set a variable for the master file (named
> e.g. `TeX-transient-master' or `TeX-session-master') the presence of
> which will be tested in `TeX-master-file' and used for `TeX-master'
> if present.
>
> In the case of shared files being opened during a `preview-document'
> call, the function `preview-parse-messages' may open files
> programmatically.  That means somewhere down the line from
> `preview-document' to that point a transient master would have to be
> set.

Probably.  I just "fixed" this stuff in the error handling of AUCTeX
so far because it terribly interfered with debugging.  But of course
the handling of preview-latex is pretty much the same here.

>> Of course, I have not actually announced this general strategy at
>> all, so I can hardly fault you for presenting a different strategy,
>> one that we previously agreed on.
>
> That's okay.  I am always grateful if somebody suggests a solution
> more convenient or beneficial for the user.  You are right that it
> would be foolish to prompt the user for information we can deduct
> from a processing session.

Yes and no.  _If_ we are going to write something into the local
variables section of the file, we better make sure that it is
something which the user really wants to have permanently recorded
there.  But I think we can make do without this in many cases.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum