Re: pgbench and timestamps (bounced)

Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr> writes:
>> Hi, this entry is "Waiting on Author" and the thread was inactive for a
>> while. I see this discussion still has some open questions. Are you
>> going to continue working on it, or should I mark it as "returned with
>> feedback" until a better time?

> IMHO the proposed fix is reasonable and addresses the issue.
> I have responded to Tom's remarks about it, and it is waiting for his
> answer to my counter arguments. So ISTM that the patch is currently still
> waiting for some feedback.

It looks like my unhappiness with injecting a pthread dependency into
pgbench is going to be overtaken by events in the "option delaying
queries" thread [1]. However, by the same token there are some conflicts
between the two patchsets, and also I prefer the other thread's approach
to portability (i.e. do it honestly, not with a private portability layer
in pgbench.c). So I'm inclined to put the parts of this patch that
require pthreads on hold till that lands.

What remains that we could do now, and perhaps back-patch, is point (2)
i.e. disallow digits as the first character of a pgbench variable name.
That would be enough to "solve" the original bug report, and it does seem
like it could be back-patched, while we're certainly not going to risk
back-patching anything as portability-fraught as introducing mutexes.

regards, tom lane

[1] https://www.postgresql.org/message-id/flat/20200227180100.zyvjwzcpiokfsqm2%40alap3.anarazel.de