Re: Saving file position in mcedit

["Andrew V. Samoilov" <sav bcs zp ua>]

Pavel Roskin wrote:
> Hello!
>
> I have tried the patch for saving file position, and indeed it seems very 
> convenient.  I think it should be in 4.6.0.
>
> I'm doing minor cleanups now.  In particular, I changed
> edit_move_to_column() to be much more effective (current patch is attached
> for the curious).
>
> There is one question I want to ask.  What if we save the positions for
> the device and inode combinations, rather than for filenames?  Device and
> inode numbers are part of the stat structure, and they are unique for
> every file.  They also survive renames.  mc already calls stat(), so we
> have the required data already.
>
> On the positive side we have less manipulations with strings in the code
> (in fact, the file with positions can just be a binary) and the ability to
> rename files and whole directories without losing saved positions for all
> files.
>
> On the negative side, it is harder to edit the file with positions, e.g.  
> to hide the fact that you were editing certain files.  I don't think it's
> important.  You can always erase the whole file to hide evidence.
>
> Any other reasons to tie positions to full filenames?
Ok, let we choose Safe save or Do backups in the editor.
Now editor rename original file to someting else and then creates new 
file and stores edited file there.  New version has new dev/inode 
combination and saved position are useless.

There is one more reason to use full filenames (or even realpath) in 
editor always. If you change working directory and filename is not 
absolute then editor will save this file relative new working directory. 
 Of course, user should see names s/he give to editor.  It was the main 
inconsistensy in the Wallery Studennikov multi-screen patche as for me.


--
Regards,
Andrew V. Samoilov