bug#32452: 26.1; gnutls_try_handshake maxes out cpu retrying when server is a bit busy

[Paul Eggert]

On 2/28/22 00:59, Lars Ingebrigtsen wrote:
> /usr/bin/ld: gnutls.o: in function `gnutls_try_handshake':
> /home/larsi/src/emacs/trunk/src/gnutls.c:634: undefined reference to 
> `rpl_nanosleep'

Evidently my recent workarounds in Emacs to handle running gnulib-tool 
from a bare checkout were not sufficient. I installed the attached patch 
to up the ante; please give it a try.

It is unfortunate that emacs/admin/merge-gnulib now runs gnulib-tool 
twice from a bare checkout, as gnulib-tool is quite slow.

I think gnulib-tool needs to run twice because it builds 
emacs/lib/gnulib.mk.in before it builds emacs/m4/gnulib-comp.m4, and so 
doesn't use the newly-added emacs/m4/nanosleep.m4 to figure out the new 
X=@X@ lines that needed to be added to emacs/lib/gnulib.mk.in. I suspect 
that this is related to Emacs's using Gnu Make rather than Automake. 
However, I haven't debugged this out.

0001-Work-around-merge-gnulib-glitch-from-fresh-checkout.patch
Description: Text Data