GNU bug report logs - #12507
24.2.50; `bookmark-write-file': use `write-file', not `write-region', to get backups

Previous Next

Full log

View this message in rfc822 format

From: "Drew Adams" <drew.adams <at> oracle.com>
To: "'Thierry Volpiatto'" <thierry.volpiatto <at> gmail.com>
Cc: 'Karl Fogel' <kfogel <at> red-bean.com>, 'Stefan Monnier' <monnier <at> iro.umontreal.ca>, 12507 <at> debbugs.gnu.org
Subject: bug#12507: [debbugs-tracker] Processed: severity 12507 wishlist
Date: Sat, 29 Sep 2012 08:51:34 -0700

> >> > The question about visiting the file is still open, though.
> >>
> >> I am using now the patch I sent here yesterday and it works 
> >> really good, faster and do backups (numered) as expected.
> >> Hope it will be applied here in emacs because it DTRT.
> >> 
> >> I don't understand what is the problem with "visiting the
> >> file". See in precedent post why it is not bad visiting the
> >> file. In the special case of bookmark-write-file, it is
> >> really not the problem.
> >
> > Your question is for Stefan.  Your patch is equivalent to
> > the change I proposed originally: just replace
> > `write-region' with `write-file'.
>
> No, this is ineficient too, you write twice the same data.

How so?  What am I missing?  What part of `write-file' writes the same data
twice?  All I see in the `write-file' definition, in terms of writing, is a call
to `save-buffer'.

> The important thing is writing directly to the buffer of file.
> For the backup thing, yes it is similar, but with unneeded steps,

Steps that you seem to claim constitute an additional disk write.  I don't see
that.  What part of `write-file' performs an extra disk write?

The only "extra" steps I see in `write-file' are setting the visited file name,
setting the buffer status to modified, checking that the file is
`file-writable-p', and setting `buffer-read-only' to nil.  And running
`vc-find-file-hook'.

You are, I think, side-tracking the issue a bit.  The question to be decided is
whether to allow backups.  It is not whether to use `write-file', `save-buffer',
`basic-save-buffer', or something else.  I don't really care exactly how it's
done.  I have confidence it will be done efficiently if it is decided to be
done.

> going straight to save-buffer is better and faster IMO (of 
> course if you have started writing data in the file buffer)
> 
> But the worst thing is the actual version with write-region:
> Slow and backup broken.

I don't see that the current version is slow, either.  But it certainly does not
provide for backing up.  That is the question to be decided.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.