Re: [developer] 16636 want ptsname_r

[Robert Mustacchi <rm@fingolfin.org>]

On 7/3/24 21:22, Yuri Pankov wrote:
> Robert Mustacchi wrote:
>> https://code.illumos.org/c/illumos-gate/+/3571
>>
>> This adds a new POSIX-specified re-entrant version of ptsname().
> 
> Hi Robert,
> 
> Is there any published document for this (a draft?)?  All I could find
> is documentation in FreeBSD and Linux where "non-standard" ptsname_r()
> is setting errno in addition to returning non-0 value, and that doesn't
> seem to match what's in review, so the standard has probably changed that?

Yes, POSIX 2024 was published. I didn't want to work off a non-draft.
The web form of it seems to not have been updated by the open group yet,
but the spec is available at
https://ieeexplore.ieee.org/document/10555529, but not for free
unfortunately. There are versions of what led to it at
https://austingroupbugs.net/view.php?id=508.

I have a "If successful, the ptsname_r( ) function shall return zero.
Otherwise, an error number shall be returned to indicate the error."
with the rationale being "Historically, some versions of ptsname( ) did
not set errno even when returning a null pointer. However, ptsname_r( )
is required to either populate the buffer with a valid name or return an
error value."

Spot checking the manual page I have for ptsname_r() in Ubuntu 22.04,
this matches. Same with musl. The FreeBSD manual page is different, but
notes it's changing for POSIX compat. Does that help at all?

Robert