* replace all occurences of `errors.Fatal(err.Error())` with `errors.Fatalf("%s", err)` so that the error wrapping is correct across the codebase
* updated the review comments
PR #5358 reintroduced a version of the TIOCGPGRP ioctl call that works
on all Unix platforms except Linux, due to a bug/inconsistency in
x/sys/unix. This commit fixes that by introducing termstatus.Tcgetpgrp.
It also introduces termstatus.Getpgrp and termstatus.Tcsetpgrp to deal
with the different signature of unix.Getpgrp in Solaris vs. all other
Unix platforms and an int-overflowing constant on AIX, so that some
AIX/Solaris-specific code can be removed elsewhere and
foreground/background detection is done the same everywhere except on
Windows.
The rclone and sftp backends require starting a child process. The are
first moved into the foreground and back into the background after the
initial startup is done.
However, this behavior is also active if restic itself is started in the
background. In this case, restic changing the foreground process may
confuse the shell and in case of bash causes it to exit. Thus, disable
modification of the controlling process group of the tty if restic is
run in the background.
this could crash in two cases:
- if a directory is deleted between restic stating it and trying to list
its directory content.
- when restic tries to list the parent directory of a backup target, but
the parent directory has been deleted.
return an error in this case instead.
This change classifies HTTP error 507 (Insufficient Storage) as a
permanent error that should not be retried. I keep running into
this once in a while and there is literally no point in retrying when
the server is full.
Fixes#5429
Signed-off-by: Dominik Schulz <dominik.schulz@gauner.org>
This adds proper support for filenames that include directories. For
example, `/foo/bar` would result in an error when trying to open `/foo`.
The directory tree is now build upfront. This ensures let's the
directory tree construction be handled only once. All accessors then
only have to look up the constructed directory entries.
The timeout for all blobs starts to run after the delete calls have been
issue. Thus, use the same start time for all blobs instead of individual
timeouts.
