Re: Variable-length FunctionCallInfoData

On 1/26/19 11:20 PM, Andrew Dunstan wrote:
> On 1/26/19 9:46 PM, Tom Lane wrote:
>> Andres Freund <andres(at)anarazel(dot)de> writes:
>>> The buildfarm largely seems to have had no problem with it, but handfish
>>> failed:
>>> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=handfish&dt=2019-01-26%2022%3A57%3A19
>>> but I have no idea what the error is, nor whether it's related to this
>>> failure, because for reasons I do not understand, the stage message is
>>> effectively empty.
>> Yeah, we've been seeing random failures with empty stage logs on multiple
>> machines. I suspect something weird with the buildfarm client script,
>> but no idea what.
>
> It's happening on a handful of animals (see below) , and at that
> intermittently. Some of the animals are leading or bleeding edge
> distros. Without more info I'm inclined to suspect environmental factors.
>

I think I have discovered the issue - small logic bug on my part :-(
Luckily it won't affect most users, especially if they are using
run_branches.pl.

Fixed in
<https://github.com/PGBuildFarm/client-code/commit/d4dd30e3b43981b51de8f5b572f131c759737813>
There will be a new release this week which will contain the fix.

cheers

andrew

--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services