Busted includes somewhere near worker_internal.h
headerscheck and cpluspluscheck are both unhappy about this:
./src/include/replication/worker_internal.h:49:2: error: unknown type name 'slock_t'
slock_t relmutex;
^~~~~~~
Now, worker_internal.h itself hasn't changed in some time.
I conclude that somebody rearranged one of the header files
it depends on. Anyone have an idea what the relevant change
was? Should we just include spin.h here, or is there a
better fix?
regards, tom lane
Hi,
On 2020-07-17 16:09:14 -0400, Tom Lane wrote:
headerscheck and cpluspluscheck are both unhappy about this:
./src/include/replication/worker_internal.h:49:2: error: unknown type name 'slock_t'
slock_t relmutex;
^~~~~~~Now, worker_internal.h itself hasn't changed in some time.
I conclude that somebody rearranged one of the header files
it depends on. Anyone have an idea what the relevant change
was? Should we just include spin.h here, or is there a
better fix?
I'm probably to blame for that - I've removed the s_lock.h (it wasn't
spin.h for some reason) include from lwlock.h:
commit f219167910ad33dfd8f1b0bba15323d71a91c4e9
Author: Andres Freund <andres@anarazel.de>
Date: 2020-06-18 19:40:09 -0700
Clean up includes of s_lock.h.
Users of spinlocks should use spin.h, not s_lock.h. And lwlock.h
hasn't utilized spinlocks for quite a while.
Discussion: /messages/by-id/20200618183041.upyrd25eosecyf3x@alap3.anarazel.de
I think including spin.h is the right fix, given that it needs to know
the size of s_lock.
Greetings,
Andres Freund
Andres Freund <andres@anarazel.de> writes:
On 2020-07-17 16:09:14 -0400, Tom Lane wrote:
headerscheck and cpluspluscheck are both unhappy about this:
./src/include/replication/worker_internal.h:49:2: error: unknown type name 'slock_t'
slock_t relmutex;
^~~~~~~
I think including spin.h is the right fix, given that it needs to know
the size of s_lock.
Done that way.
regards, tom lane