Le 20/06/2011 00:49, Lex Trotman a écrit :
Just tell me -- what's the advantages of 2 over 3 then? OK, it wouldn't depend on gvfs-fuse.
Yes, and it is safe (I think) if it keeps the backup file.
True. And it would be safe against write failures with GEdit's hack or similar.
[...]
PS just for info last time I looked Gedit used option 2 but with special implementation dependent hacking to make it safe. Maybe stealing that is the best option, then its biggest downside is being slow sometimes.
Yes, we could make GIO saving safe (e.g. handle writes failure) by haking on until #602412 [1] get fixed properly. This looks kinda ugly, but would at least provide safe saving, just leaving out the temporary file on failure.
My understanding is that the nasty hack in Gedit is to work around this problem.
Yes it is. I also wrote an example a while ago with a similar hack, and it worked -- will need to check which hack is the best, but still a hack...
But it wouldn't address the cases where GIO won't success to properly write (e.g. the bugs I linked in previous mail). Maybe it is fixable in GIO though, so reporting those to them might be a start point -- though I can't tell whether it's reproducible in recent GIO versions, nor what kind of problem it is.
3220784 doesn't say if safe file saving is set (ie its using type1), its the default and doesn't use GIO but no one asked the OP also notes gedit works so stealing their hack would work.
Right. Simply using GIO would work without the hack I guess, the hack just prevents data loss upon write failure.
3184594 is using GIO, the temporary file is a GIO name. The question is what sort of filesystem does virtual box report the host to the guest as, this may confuse GIO.
Yeah...
3126535 doesn't say if safe file saving is set (ie its using type1), its the default and doesn't use GIO but no one asked the OP
I guess she's using GIO (.goutputstream-KMEUMV is a GIO name), and and he seems to have the same problem with GEdit. Looks like the target file was still open ("Text file busy"), so it's either a GIO problem (failed to detect it needs closing the file on a Windows share), an OP's problem (she has the file opened on the remote side) or a sharing problem (whatever).
Le 19/06/2011 13:42, Enrico Tröger a écrit :
On Tue, 14 Jun 2011 10:28:00 +1000, Lex wrote:
So I would say that all three need to stay, and the choice between >>
GIO and the c library needs to be made a preference not a compile
time option.
I think so too. I'm only afraid how to give this choice to the user. This needs to be well documented and probably should go into a normal. not hidden pref and so should be in the preferences dialog.
Agreed. However, I can't see how to present this to the user in a way she can take a proper decision... In the meantime, I think I can commit the patch I proposed -- we will still be able to make it a visible pref later -- nope? Though, maybe it'll make us lazy to do the change... :D
I agree that it is important enough to add the patch as an interim until the UI or some other fix can be worked out.
Committed.
Cheers, Colomban