[Geany-Devel] [Geany-devel] Separating session file lists from config (again)

[Thomas Martitz]

Am 19.11.2013 00:14, schrieb Lex Trotman:
> [...]
>
>
>     >> 1. Project and user session file formats should match to allow the
>     >> same code to store both.  The fact that the storing the current
>     >> project in the project session file is not needed doesn't matter.
>
>     It looks like a good idea, I agree, but we need to keep on allowing to
>     save user-wide settings when a project is open.  I mean, we shouldn't
>     save all settings changes to the project settings just because one is
>     open, the user might want to change a general setting.
>
>     I think this was discussed one day already, and I think it should be a
>     separate thread.
>
>
> I said **session** file format, not the config files, as you say ATM 
> the only change to config should be to remove the session info.  All 
> further changes are indeed another topic.
>
> [...]
>
>     >
>     > Old Geany couldn't read it since it's a directory, and I don't think
>     > you can possibly overwrite a directory with a file, even
>     > accidentally. You'd need to delete the directory first.
>
>     I agree, I don't think this should be an issue.  Older Geany would not
>     be able to load the project, but that should be all.  And if the
>     config
>     part starts with the same format as old $project.geany, a user could
>     workaround that by loading $project.geany/project.conf manually.
>
>
> I'm happy for it to be the same so long as someone can assure me that 
> geany saving a file can't overwrite a directory on the most common 
> file system arrangements which includes windows and includes windows 
> file systems on a NAS mounted via Samba (a sadly increasing situation 
> with cheep NAS boxes).
>
> Note that config saving will use g_file_set_contents() if the option 
> is set, which writes to a temp file, and renames it over the old file. 
>  I agree that on Linux it can't rename a file over a directory, but on 
> what it thinks is a windows file system it deletes the old file first 
> and then renames the temp to it.  Are we sure that won't allow it to 
> overwrite directories?
>
>


I would consider it a bug in g_file_set_contents() if doesn't check 
permissions and type of the original object (file) before doing the 
temp&rename approach. Because that would mean different semantics 
compared to direct save.

FWIW, let's not overcomplicate this. We can easily chose another naming 
scheme for the directory.

Best regards.