Re: [R21.4] buffer-file-truename vs. info.el
"stephen" == stephen <stephen(a)xemacs.org> escreveu:
stephen> RECOMMEND 21.4
stephen> Vin, please review. The recommendation is based on theory;
stephen> I will be testing, and Nelson too, I guess.
I am sorry, but should there have been a patch attached?
stephen> A separate patch will be submitted for 21.5, because 21.4 setqs
stephen> Info-index-alternatives nil, but in 21.5 it's commented out:
[...]
stephen> Nelson Ferreira writes:
> This happens for me mostly when using Info and the info file is
> compressed. I believe the root cause of this is because the buffer is
> not associated with a file name.
stephen> Not quite right. It is associated with a file name, but that code
stephen> that does it was faulty.
Ok.
> The issue here seems to be that on the decompressed bufefr
> buffer-file-truename is nil and while buffer-file-name still
> holds the .gz file name which can be used to save the font-lock
> info, but it was not.
stephen> Aha! Functions like `find-file' are supposed to maintain
stephen> the invariant `(equal buffer-file-truename (file-truename
stephen> buffer-file-name))', but info.el failed to do so. The
stephen> following patch fixes that bug. Will you check and see if
stephen> it fixes the symptom for you?
Sure, I will as soon as I get the patch :)
stephen> Would you also review your experience briefly, and let me
stephen> know if you can think of any other situation where this has
stephen> happened? I briefly audited the Lisp code, which seems OK,
stephen> but the C code is much harder to review.
This has also happened to me in some ERC buffers, specially the ones
where part of the history from logs were pulled out from the disk. I
will try to remember if there were any more cases. Those 2 are
certainly the ones I encounter repeatedly.
--
Nelson Ferreira
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches