On Tue, 2023-02-21 at 08:35 -0500, Ken Brown wrote: > On 2/21/2023 3:54 AM, Ken Brown wrote: > > On 2/21/2023 12:27 AM, Paul Smith wrote: > > > Just to note, I do run the regression test suite with the > > > equivalent of "jobserver-style=pipe" (basically I force the > > > configure to believe that the system doesn't support mkfifo() so > > > it falls back to "pipe" mode). > > Somehow the test suite isn't catching this problem, at least on > Cygwin.
Understood. I'm just saying it's not the case that there is no testing at all with "pipe", anymore. Agreed that the regression tests are not completely sufficient to find all problems. > make[5]: not recursive -2,-2 > make[5]: warning: jobserver unavailable: using -j1. Add '+' to > parent make rule. Perfect, thanks. That probably means the message is the result of an explicit change we made to lock down the pipe mode, rather than some unanticipated modification. I'll look at the makefile you provided to see if I can figure it out. Maybe the changes we made need to be adjusted. For more info, see this entry in the NEWS file: * When the pipe-based jobserver is enabled and GNU Make decides it is invoking a non-make sub-process and closes the jobserver pipes, it will now add a new option to the MAKEFLAGS environment variable that disables the jobserver. This prevents sub-processes that invoke make from accidentally using other open file descriptors as jobserver pipes. For more information see https://savannah.gnu.org/bugs/?57242 and https://savannah.gnu.org/bugs/?62397 The "option that disables the jobserver" is where the -2,-2 comes from.