On Tue, Jul 15, 2025 at 08:34:52PM +0200, Ingo Schwarze wrote:
> Hello Angelo and Jonathan,
> 
> Angelo Rossi wrote on Tue, Jul 15, 2025 at 05:44:19PM +0200:
> 
> > In OpenBSD 7.7 the manual page for errno does not report the error 71:
> > EREMOTE --- 71 --- Too many levels of remote in path.
> 
> Unless i'm misreading the code of our system, the only driver
> generating that particular error number is inteldrm(4), and even
> there my impression is it was only added in February 2025 by jsg@,
> and only in sys/dev/pci/drm/i915/display/intel_fb_bo.c.
> The C library does not generate this error number.
> 
> My impression is this error number can only get set as a result of
> ioctl(2) against an inteldrm(4) device.  In that region, i think
> we have have significantly worse documentation gaps: The inteldrm(4)
> manual doesn't even mention at all that inteldrm(4) supports ioctl(2),
> and it does not list which ioctls it supports, even though that is
> what driver manual pages are supposed to do when the driver supports
> driver-specific ioctls.  After ioctls are documented, one can then
> add an ERRORS section describing under which conditions the
> device-specific ioctls fail.
> 
> I'm not opposed to adding EREMOTE to errno(2), but as long as the
> command "man -k Er=EREMOTE" would return only a single page, errno(2),
> the benefit for users would seem quite limited to me, in particular
> with respect to an error message as cryptic as "Too many levels of
> remote in path" - what is a user supposed to learn from that message
> as long as they cannot even figure out that it's related to trying
> to configure a framebuffer for inteldrm(4), which seems perfectly
> obscure to me given the wording of the error message?
> 
> Jonathan, what do you think should be done here?
>   Ingo

EREMOTE use in inteldrm comes from linux, not somthing I changed.
inteldrm is not a reason to document it, I don't think that codepath
is even reachable at the moment.