document-portal: Don't invalidate directories #1243
Conversation
hfiguiere
force-pushed
the
issue-1234
branch
2 times, most recently
from
c7bd171
to
1ba8834
Compare
|
Turns out it prevents the filesystem to be ejected. So this is unlikely to be the right fix. |
|
Right now the situation is as follow:
I haven't found a different solution that would not hold the directories too long. |
This works around issue flatpak#1234. This is a followup from flatpak#1190. fuse_lowlevel_notify_inval_entry() will invalidate directories too, which is a problem as it invalidate the current workdir directory. So now we don't invalidate them and will let the kernel deal with it. This is consistent with the passthrough_hp example from fuse. But this still make the directory being held for longer than necessary. Signed-off-by: Hubert Figuière <[email protected]>
|
Does it even make sense for anything to be using a directory under (A little rant about applications using `/run/user/1000/doc/`)While on this topic, something that has been bothering me a lot is that if I save a file in a flatpak Chromium that uses xdg-desktop-portal to a remote drive at |
This work around issue #1234
fuse_lowlevel_notify_inval_entry() will invalidate directories too, which is a problem as it invalidate the current workdir directory. So now we don't invalidate them and will let the kernel deal with it.
This is consistent with the passthrough_hp example from fuse.