Opened 2 months ago
Last modified 2 months ago
#19048 new bug
*at() methods accept -1 for AT_FDCWD
Reported by: | bhaible | Owned by: | nobody |
---|---|---|---|
Priority: | normal | Milestone: | Unscheduled |
Component: | System/POSIX | Version: | R1/beta5 |
Keywords: | Cc: | ||
Blocked By: | Blocking: | #19049, #19050, #19051, #19052, #19053, #19054, #19055, #19056, #19057, #19058, #19059, #19061 | |
Platform: | All |
Description
POSIX https://pubs.opengroup.org/onlinepubs/9799919799/functions/readlink.html specifies that the readlinkat() function "shall fail" with error code EBADF if "The path argument does not specify an absolute path and the fd argument is neither AT_FDCWD nor a valid file descriptor open for reading or searching."
This does not work in Haiku hrev57823 (from 2024-07-15): it fails with error code ENOENT instead.
How to reproduce: Compile and run the attached test program.
gcc -Wall foo.c ./a.out
Expected output:
ret = -1, errno == EBADF OK
Actual output:
ret = -1, errno == No such file or directory a.out: foo.c:20:main: ret < 0 && errno == EBADF Kill Thread
Attachments (1)
Change History (16)
by , 2 months ago
comment:1 by , 2 months ago
This, and all the other *at() tickets you have opened today, all have one and the same cause: -1 used to be the value for AT_FDCWD on Haiku. This was changed to -100 so things are clearer, but the kernel still accepts -1 so we don't break ABI too soon. (BeOS did not have AT_FDCWD, so we can change this without breaking ABI with BeOS.)
If one uses a large positive but nonexistent FD, then the tests all work as expected.
comment:2 by , 2 months ago
If one uses a large positive but nonexistent FD, then the tests all work as expected.
I confirm this part: I tried fd = 99, and got EBADF in this case.
comment:3 by , 2 months ago
Summary: | readlinkat() does not fail with EBADF when it should → *at() methods accept -1 for AT_FDCWD |
---|
I've adjusted things in hrev58059 so that the kernel at least rejects anything which isn't -1 or AT_FDCWD, so -2 and other such values should now trigger the error.
Changing this ticket's title and will consolidate the other tickets into it.
comment:4 by , 2 months ago
Blocking: | 19049 added |
---|
comment:5 by , 2 months ago
Blocking: | 19050 added |
---|
comment:6 by , 2 months ago
Blocking: | 19051 added |
---|
comment:7 by , 2 months ago
Blocking: | 19052 added |
---|
comment:8 by , 2 months ago
Blocking: | 19053 added |
---|
comment:9 by , 2 months ago
Blocking: | 19054 added |
---|
comment:10 by , 2 months ago
Blocking: | 19055 added |
---|
comment:11 by , 2 months ago
Blocking: | 19056 added |
---|
comment:12 by , 2 months ago
Blocking: | 19057 added |
---|
comment:13 by , 2 months ago
Blocking: | 19058 added |
---|
comment:14 by , 2 months ago
Blocking: | 19059 added |
---|
comment:15 by , 2 months ago
Blocking: | 19061 added |
---|
test case foo.c