17.10.2020 21:44, Tom Lane wrote:
> I propose the attached patch. If this doesn't cause buildfarm problems,
> perhaps we should back-patch it.
Thank you!
I've made a simple cmd script to reproduce problems seen on dory:
https://buildfarm.postgresql.org/cgi-bin/show_history.pl?nm=dory&br=HEAD
FOR /L %%I IN (1,1,200) DO call :CHECK %%I
GOTO :eof
:CHECK
echo iteration %1
call vcregress ecpgcheck
IF %ERRORLEVEL% NEQ 0 GOTO ERR
EXIT /B
:ERR
echo iteration %1 failed
pause
Without the fix I've got errors on iterations 43, 46, 128, 47, 14, 4,
27, which approximately corresponds to the ECPG-Check failure frequency
on dory (for HEAD).
With the fix all the 200 iterations passed as expected.
Then I ran the loop again just to be sure and got:
test thread/descriptor ... stderr FAILED 81 ms
iteration 124 failed.
diff -w -U3
.../src/interfaces/ecpg/test/expected/thread-descriptor.stderr
.../src/interfaces/ecpg/test/results/thread-descriptor.stderr
--- .../src/interfaces/ecpg/test/expected/thread-descriptor.stderr
2019-12-04 16:05:46 +0300
+++ .../src/interfaces/ecpg/test/results/thread-descriptor.stderr
2020-10-18 20:20:27 +0300
@@ -0,0 +1 @@
+SQL error: descriptor "mydesc" not found on line 31
It's interesting that all failures before the fix were with stdout, but
this one is with stderr.
I'm going to investigate this issue further.
Best regards,
Alexander