x/sys/unix: Linux: utimensat syscall takes a NULL path, wrapper can't

[unbrice]

What version of Go are you using (go version)?

x/sys package v0.0.0-20211117180635-dee7805ff2e1

Does this issue reproduce with the latest release?

Yes.

What operating system and processor architecture are you using (go env)?

go env Output

The version of Golang we use at Google (I believe this is not public information, please correct me if I'm wrong and I'll share go env)

What did you do?

I opened the documentation of sys/unix and also read its source code.

What did you expect to see?

I expected to be able to do the equivalent to the futimens glibc function.
This could have been its own function. Alternatively the wrapper of utimensat could have given empty strings the special meaning that NULL pointers have for the syscall.

(On Linux, in the glibc, futimens is implemented on top of utimensat by passing a NULL pointer as path. This is described in the NOTES section of the manpage.)

I also expected to see Futimes implemented in a way that did not go through /proc

What did you see instead?

• The wrapper of utimensat, named UtimesNanoAt, takes a string for the path and does not have a special case to pass NULL when the string is empty.
• The wrapper of Utimes goes to /proc instead of taking advantage of a NULL path
• There is no equivalent for futimens

Please let me know if you'd like me to contribute a fix. Unless someone suggest another approach, I offer to send a PR that leaves UtimesNanoAt unchanged and adds FUtimesNano.

[ianlancetaylor]

When I read the man page of the utimensat I don't see any difference between a NULL pathname argument as opposed to a pathname argument that is the empty string. When is it important to pass NULL, rather than a pointer to a NUL byte, to utimensat?

[unbrice]

As per the NOTES section of the manpage, the difference is as follow;

• a NULL path results in the call being applied to the file referred to by the file descriptor dirfd (which does not need to be a directory). This allows implementing futimens/FUtimesNano.
• an empty path returns the ENOENT error.

[ianlancetaylor]

Thanks. I guess we should consider adding a new syscall wrapper that passes NULL. Perhaps this should be Futimens.