| Commit message (Collapse) | Author | Age |
|
|
|
|
| |
Ensure that we are really creating a new file.
Don't attempt to write, we do that in tst_open_write().
|
|
|
|
|
| |
We are actually testing both opening of an existing file
and writing to it.
|
|
|
|
| |
To check for unlink() support without requiring create()/mknod().
|
|
|
|
|
| |
This allows testing a filesystem that offers mkdir(), but no
rmdir() (and vice versa).
|
|
|
|
|
| |
This makes more sense, since we are specifically checking
unlinking of an open file.
|
|
|
|
|
|
| |
By creating the files in the lower filesystem, we
can test readdir() even for filesystems that don't implement
create() or mkdir().
|
| |
|
| |
|
|
|
|
| |
See also issue #148.
|
|
|
|
| |
No longer supported in Meson 0.39.
|
| |
|
|
|
|
|
| |
The FUSE_CAP_ATOMIC_IO_TRUNC capability is enabled by default,
but we didn't update the open() documentation accordingly.
|
|
|
|
| |
Fixes #138.
|
|
|
| |
Redundant copy when only op.read is available removed.
|
|
|
|
| |
Putting it in CFLAGS interferes with feature detection.
|
|
|
|
|
| |
Defining it in the file causes trouble because Meson sometimes
inserts includes before the first line.
|
| |
|
| |
|
| |
|
|
|
|
| |
This triggered undefined behaviour warnings from UBSan.
|
|
|
|
|
|
|
|
|
|
|
|
| |
The permission caching bug has been present forever, is presumably
going to stay around for a while, and is of less concern if
allow_other is not used. Since allow_other is disabled by default, I
think we can safely make this warning less prominent and document the
problem when we describe allow_other.
Also, drop the travis build status. It's confusing when reading
README.md after extracting the tarball, and I am not sure who benefits
from the build status when it is shown on GitHub either.
|
| |
|
| |
|
|
|
|
|
| |
Since existing whitespace usage is not consistent, this causes
a lot of spurious whitespace changes.
|
|
|
|
|
|
|
| |
Apparently, NetBSD does not have /etc/mtab. Setting IGNORE_MTAB
in this case makes the code a little nicer.
See also https://github.com/libfuse/libfuse/pull/123
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This reverts commit d5cdbb94a0650b0a462682cf0a84463ff1513900.
null works completely fine, just the mountpoint should
be a regular file -- so there is no need to dismiss it.
Also:
- fixing up compiler warnings
- checking if the mountpoint is a regular file
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
documentation
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Otherwise building w/o optimization (-O0) fails.
passthrough_ll.c: In function 'lo_readlink':
passthrough_ll.c:251:11: error: 'PATH_MAX' undeclared (first use in this function)
char buf[PATH_MAX + 1];
(gcc v5.3.0 in Alpine Linux v3.4.6 x86_64 w/ musl-libc v1.1.14.)
|
|
|
|
|
| |
If we want to do, what system would do anyway upon program termination,
then let's try to be a bit more scrupulous.
|
|
|
|
|
|
| |
root lo_inode's next and prev were not set early enough, which led to
accessing addr 0x8 (or 0x4, depending on ptr size) when setting prev
variable in lo_free(lo.root.next), because lo.root.next was NULL.
|
| |
|
| |
|
|
|
|
|
|
| |
-oallow_root is handled in userspace, and requires passing -oallow_other
to the kernel. This patch should make the code easier to understand and
avoid the confusion that gave rise to issue #86.
|
|
|
|
| |
Fixes #116.
|
|
|
|
| |
Fixes #117.
|
|
|
|
|
|
| |
Enabled by default since we haven't released libfuse 3.0 yet :-).
Fixes #112.
|
|
|
|
| |
Taken from Linux kernel commit 27bcd37.
|
|
|
|
| |
Fixes #95.
|
| |
|
| |
|