file-precious-flag not taken seriously enough?

file-precious-flag not taken seriously enough?

[Karl Fogel]

In lisp/files.el:(basic-save-buffer-2), we have this code:

    (let ((dir (file-name-directory buffer-file-name)))
      (if (and file-precious-flag
	       (file-writable-p dir))
	  ;; If file is precious, write temp name, then rename it.
	  ;; This requires write access to the containing dir,
	  ;; which is why we don't try it if we don't have that access.
	  (let ((realname buffer-file-name)
		tempname succeed
		(umask (default-file-modes))
		(old-modtime (visited-file-modtime)))
	[...continue on to write directly to the file...]

The documentation for `file-precious-flag' doesn't say anything about
it only applying if the containing directory is writeable.  But that
seems to be how we treat it, in the conditional above.

Is this just a bug?

(Larger context: I'm implementing the `break-hardlinks-on-save' flag
as discussed in another thread, so am looking at where
`file-precious-flag' is used.)

-Karl

Re: file-precious-flag not taken seriously enough?

[Chong Yidong]

Karl Fogel <kfogel@red-bean.com> writes:

>     (let ((dir (file-name-directory buffer-file-name)))
>       (if (and file-precious-flag
> 	       (file-writable-p dir))
> 	  ;; If file is precious, write temp name, then rename it.
> 	  ;; This requires write access to the containing dir,
> 	  ;; which is why we don't try it if we don't have that access.
> 	  (let ((realname buffer-file-name)
> 		tempname succeed
> 		(umask (default-file-modes))
> 		(old-modtime (visited-file-modtime)))
> 	[...continue on to write directly to the file...]
>
> The documentation for `file-precious-flag' doesn't say anything about
> it only applying if the containing directory is writeable.  But that
> seems to be how we treat it, in the conditional above.
>
> Is this just a bug?

If the directory is not writable, there is no way to obey
file-precious-flag.  That means that the only alternative is to signal
an error.  Off the top of my head, I think it's better to fall back on
the usual file saving method instead, because that's almost always what
the user would want.

Re: file-precious-flag not taken seriously enough?

[Kevin Rodgers]

Chong Yidong wrote:
> Karl Fogel <kfogel@red-bean.com> writes:
> 
>>     (let ((dir (file-name-directory buffer-file-name)))
>>       (if (and file-precious-flag
>> 	       (file-writable-p dir))
>> 	  ;; If file is precious, write temp name, then rename it.
>> 	  ;; This requires write access to the containing dir,
>> 	  ;; which is why we don't try it if we don't have that access.
>> 	  (let ((realname buffer-file-name)
>> 		tempname succeed
>> 		(umask (default-file-modes))
>> 		(old-modtime (visited-file-modtime)))
>> 	[...continue on to write directly to the file...]
>>
>> The documentation for `file-precious-flag' doesn't say anything about
>> it only applying if the containing directory is writeable.  But that
>> seems to be how we treat it, in the conditional above.
>>
>> Is this just a bug?
> 
> If the directory is not writable, there is no way to obey
> file-precious-flag.  That means that the only alternative is to signal
> an error.  Off the top of my head, I think it's better to fall back on
> the usual file saving method instead, because that's almost always what
> the user would want.

In case the usual file saving method results in a problem (which
file-precious-flag was set to prevent), shouldn't there be a warning
message here to inform the user?  Not to distract him/her, but so that
they can find a clue in the *Messages* buffer after something has gone
wrong.

-- 
Kevin Rodgers
Denver, Colorado, USA

Re: file-precious-flag not taken seriously enough?

[Karl Fogel]

Kevin Rodgers <kevin.d.rodgers@gmail.com> writes:
> In case the usual file saving method results in a problem (which
> file-precious-flag was set to prevent), shouldn't there be a warning
> message here to inform the user?  Not to distract him/her, but so that
> they can find a clue in the *Messages* buffer after something has gone
> wrong.

Or another option: at least flash a message that file-precious-flag is
going to be ignored, when that's the case.

However, for now I'm just going to change the file-precious-flag
documentation to indicate that it is advisory.  This was all just stuff
I found on the way to making another change, which I'd still like to
make :-).

Re: file-precious-flag not taken seriously enough?

[Stefan Monnier]

>> In case the usual file saving method results in a problem (which
>> file-precious-flag was set to prevent), shouldn't there be a warning
>> message here to inform the user?  Not to distract him/her, but so that
>> they can find a clue in the *Messages* buffer after something has gone
>> wrong.
> Or another option: at least flash a message that file-precious-flag is
> going to be ignored, when that's the case.

Yes, a message like "Dir read-only: ignoring file-precious-flag" sounds like
a good idea.


        Stefan