On Mar 14 21:25, Achim Gratz wrote:
> Achim Gratz writes:
> > That was an exercise in futility.  Procps uses system macros to
> > dissemble the major/minor devices, so of course it's doing the right
> > thing on Cygwin while using Cygwin macros (I think these are actually
> > inline functions now).
> 
> Actually, it didn't -- but I patched it to do so now.
> 
> We should think a bit about whether or not Cygwin should conform to the
> /proc/…/stat definition of Linux with it's peculiar bit allotment for
> the major/minor device numbers, in which case I'll have to change to a
> versioned #ifdef (I don't particularly want to use a runtime version
> check).

Here's my stance:

- We can keep the code as is, or change that to Linux-compat, whatever
  you think is the right thing for procps-ng.

- Whatever you decide, the result will go into the next Cygwin version.

- When I release the next Cygwin version, you update procps-ng and
  there won't be any reason to be backward compat.

Ideally: Decide today, and Cygwin 3.0.4 will be ready and released this
weekend.

> > But your assertion that it should use stat for
> > finding the tty is probably wrong, there's Cygwin specific code that
> > looks at /dev/ctty:
> >
> > https://gitlab.com/procps-ng/procps/blob/master/proc/devname.c#L303
> >
> > That's used in addition to code that would correctly determine the
> > devices fromt the stat info, so at the moment I have no idea where the
> > wheels fall off that wagon.
> 
> I still don't know where the bug is in the current package, as just
> building the current version (without my patch that removes the function
> that uses /proc/ctty) also works correctly.

¯\_(ツ)_/¯


Corinna

-- 
Corinna Vinschen
Cygwin Maintainer