posix_fallocate

I would like to add "posix_fallocate" test based on what we already have in syscalls/fallocate, because the behavior of "posix_fallocate" is close to what is doing fallocate by default, so most test scenarios will be the same. What is the correct way to do it, place it in open_posix_testsuite as new ones or add it to fallocate tests?

Quoted 8 lines

That is hard to tell.

If you care for posix_fallocate strictly POSIX conformace add it to open_posix_testsuite and write the tests accordingly to the POSIX specification.

If you care about API coverage add it to the syscalls testcases and share as much code with the fallocate testcases as possible.

OK, I will add it to syscalls. What I would want is to make a wrapper, then substitute all calls of fallocate with it. And decision which function to call decided inside wrapper using command line option. So they would be called this way:

...
fallocate01 fallocate01
fallocate02 fallocate02
fallocate03 fallocate03

posix_fallocate01 fallocate01 -x
posix_fallocate02 fallocate02 -x
posix_fallocate03 fallocate03 -x
...

Quoted 17 lines

Quoted 10 lines

OK, I will add it to syscalls. What I would want is to make a wrapper, then substitute all calls of fallocate with it. And decision which function to call decided inside wrapper using command line option. So they would be called this way:

...
fallocate01 fallocate01
fallocate02 fallocate02
fallocate03 fallocate03

Quoted 4 lines

posix_fallocate01 fallocate01 -x
posix_fallocate02 fallocate02 -x
posix_fallocate03 fallocate03 -x
...

Or you can recompile the the binaries with -DUSE_POSIX_FALLOCATE and use #ifdefs to choose which do you call.

Quoted 2 lines

Looking at the testcases you will probable need to ifdef a few test assertions (these where mode is not 0) too.

Just in one place in fallocate01, where it tests "FALLOC_FL_KEEP_SIZE" mode. Still it would probably be better to have a wrapper at least for posix_fallocate, e.g. to preserve the way the code checks for errors, because it returns error number and doesn't set "errno".