go to bug id or search bugs for
Filesystem functions have IMHO the wrong behaviuor on disk-full conditions
$fp = fopen('/dev/full', 'wb');
bool(false) and "warning ...No space left on device.. (aka, handle ENOSPC)
Add a Patch
Add a Pull Request
Why would fflush return false? Nothing was written by fwrite, so the flush is a no-op.
A liltte bit better test case:
# dd if=/dev/zero of=/tmp/vfs bs=1024 count=1024
# losetup /dev/loop0 /tmp/vfs
# mkfs -t ext2 -m 1 -v /dev/loop0
# mkdir /mnt/vfs
# mount -t ext2 /dev/loop0 /mnt/vfs
$fp = fopen('/mnt/vfs/foo.txt', 'wb');
var_dump(fwrite($fp, str_repeat("fail", 1024000)));
ls -l /mnt/vfs/foo.txt
-rw-r--r-- 1 root root 1003520 oct 14 21:43 /mnt/vfs/foo.txt
Partial data on disk, no warning or return values hinting the problem.
Well, there's the hint that the return value is smaller than strlen(str_repeat("fail", 1024000)) = 4096000.
I'm not sure if adding a warning here is appropriate, we try to avoid warnings in correct scripts so that programmers don't have to use "@" to have notice/warning free code. Even if we consider a disk full an exceptional circumstance that merited breaking this guideline, a warning would be of little use; unless the logs are in a separate filesystems, the warning message would not be able to be logged.
* We can't return false, because a part of the data may have been written and we need to return how much.
* A warning would be of no use in most circumstances.
Maybe we could return a false+warning if no data has been written, but it seems dangerous because sometimes programmers would be warned of an out-of-disk-space conditional and other times they wouldn't.
Closeing as Wont Fix per the exposed reasons.