Re: [HACKERS] win32 crash in initdb
The crash people are getting in initdb is due to the 1.19 revision to
pgtz.c. Not sure why yet...looking into it...Unfortunate behavior when probing pre-1970 dates, maybe?
Indeed that was the problem. Attached patch fixes the issue here and for
Merlin :-)
I've also figured out why it always worked for me. When in the western
europe timezone, the bug does not show up (probably because we don't
probe those dates). When I set my system to US Eastern, it broke.
//Magnus
Attachments:
tzfix.patchapplication/octet-stream; name=tzfix.patchDownload+4-0
"Magnus Hagander" <mha@sollentuna.net> writes:
Indeed that was the problem. Attached patch fixes the issue here and for
Merlin :-)
This is not a good patch because it will treat the predictable pre-1970
failure as meaning the TZ should be rejected entirely. What we want to
do is treat a null result as a non-match and score the TZ according to
the number of prior successful matches.
regards, tom lane
The crash people are getting in initdb is due to the 1.19
revision to
pgtz.c. Not sure why yet...looking into it...
Unfortunate behavior when probing pre-1970 dates, maybe?
Indeed that was the problem. Attached patch fixes the issue
here and for Merlin :-)I've also figured out why it always worked for me. When in
the western europe timezone, the bug does not show up
(probably because we don't probe those dates). When I set my
system to US Eastern, it broke.
Here we go again with a more correct patch, per Toms comments and hints
off-list.
//Magnus
Attachments:
tzfix.patchapplication/octet-stream; name=tzfix.patchDownload+12-0
Import Notes
Resolved by subject fallback