Thread: pgsql: Make WaitLatch's WL_POSTMASTER_DEATH result trustworthy; simplif
Make WaitLatch's WL_POSTMASTER_DEATH result trustworthy; simplify callers. Per a suggestion from Peter Geoghegan, make WaitLatch responsible for verifying that the WL_POSTMASTER_DEATH bit it returns is truthful (by testing PostmasterIsAlive). Then simplify its callers, who no longer need to do that for themselves. Remove weasel wording about falsely-set result bits from WaitLatch's API contract. Branch ------ master Details ------- http://git.postgresql.org/pg/commitdiff/f40022f1adaeff85b01d98fea38cf5aa16814aa7 Modified Files -------------- src/backend/port/unix_latch.c | 35 ++++++++++++++++++++++---------- src/backend/port/win32_latch.c | 12 +++++++++- src/backend/postmaster/autovacuum.c | 9 ++++--- src/backend/postmaster/bgwriter.c | 7 +---- src/backend/postmaster/checkpointer.c | 7 +---- src/backend/postmaster/pgstat.c | 7 +---- src/backend/postmaster/walwriter.c | 7 +---- 7 files changed, 47 insertions(+), 37 deletions(-)
Re: pgsql: Make WaitLatch's WL_POSTMASTER_DEATH result trustworthy; simplif
From
Peter Geoghegan
Date:
On 10 May 2012 19:35, Tom Lane <tgl@sss.pgh.pa.us> wrote: > Remove weasel wording about falsely-set result bits from > WaitLatch's API contract. Aren't those weasel words still applicable to the case where sock != PGINVALID_SOCKET ? -- Peter Geoghegan http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training and Services
Peter Geoghegan <peter@2ndquadrant.com> writes:
> On 10 May 2012 19:35, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Remove weasel wording about falsely-set result bits from
>> WaitLatch's API contract.
> Aren't those weasel words still applicable to the case where sock !=
> PGINVALID_SOCKET ?
I don't think so. WaitLatch is honestly reporting the result of select
or poll; code that doesn't work with it wouldn't work with the native
OS facilities either. The only case that is really a problem here is
where we make the logical leap from "socket possibly readable" to
"postmaster is dead".
I think what that Linux man page is advising is that you not assume that
a blocking-mode read() won't block, which AFAIK is something we don't
do; we always use non-block mode on sockets we don't wish to block on.
regards, tom lane