On Tue, Sep 21, 2021 at 02:52:16PM +0200, Christian Weisgerber wrote:
> Lucas:
> 
> > Find a patch providing a replacement for `jot`, which should aid in
> > running regress under Linux. While there, I converted to obvious cases
> > of `echo 1; echo 2; echo 3; ...` to the new function.
> 
> Why not just name the function jot() so we don't have to change all
> calls?

I agree, that would help reduce the number of changes.

> > I only tested this fully under OpenBSD. I tested it lightly on Linux and
> > I don't have access to a FreeBSD system and sadly vmd can't run it.
> 
> I maintain a separate FreeBSD port.
> https://cgit.freebsd.org/ports/tree/devel/got
> 
> In fact, the -portable people may want to take a look at that for
> an overview of problems I have solved there and how I did it:
> - I use sed to insert
>   #include "openbsd-compat.h"
>   into all *.[cy] files between the #include <...> and #include "..."
>   headers.

Hmm.  Well, I much prefer to keep this in code and let any merge conflicts
surface.  That allows for legitimate changes to be looked at, especially as
there's typically other changes that have happened in the same file.

> - There are various compatibility headers/wrappers.
>
> I also recommend looking at OpenSSH for a template for the whole
> autotools stuff.

I don't think there's anything missing now.

> I would happily merge the FreeBSD changes into -portable, eventually.
> There was a long break where no work at all was happening in
> -portable, and the next few weeks I'll be very short on time.

I'd welcome that.  Note that "no activity" can be read more as "wasn't
rebasing against the changes happening in got".  The framework is there and
seems to work.

So if you'd like help with that, let me know -- it sounds like a good idea to
me.

Kindly,
Thomas