(found because my "stixky" article did not show off:)
Time zone (GMT offset in seconds): Europe/Berlin (3600)
Database server time offset: 3601 s
TXP reports the GMT offset as 3600 seconds (1 hour), but the actual database server offset is 3601 seconds — meaning a 1-second discrepancy. That is exactly the difference we saw in the SQL:
Posted: 2026-05-08 08:35:32
cutoff: 2026-05-08 08:35:31 ← 1 second too early
This is a bug in TXP 4.9.1 — the...
Edit by @petecooper to append translated details from #2074 (comment)
This is a bug in TXP 4.9.1—the offset is calculated in whole hours (3600), but the actual database offset shows a 1-second discrepancy due to Percona Server 8.4.7.
Immediate workaround: When creating a new article, always set the "Posted" date to at least one minute in the past—this reliably bypasses the issue.
Permanent solution: Report this bug on GitHub using exactly this diagnostic data:
👉 https://github.com/textpattern/textpattern/issues
The diagnostic output you have is perfect for a bug report— "Database server time offset: 3601 s" is the key piece of evidence. It should actually be 3600, and TXP ought to handle this more robustly.
(found because my "stixky" article did not show off:)
TXP reports the GMT offset as 3600 seconds (1 hour), but the actual database server offset is 3601 seconds — meaning a 1-second discrepancy. That is exactly the difference we saw in the SQL:
This is a bug in TXP 4.9.1 — the...
Edit by @petecooper to append translated details from #2074 (comment)
This is a bug in TXP 4.9.1—the offset is calculated in whole hours (3600), but the actual database offset shows a 1-second discrepancy due to Percona Server 8.4.7.
Immediate workaround: When creating a new article, always set the "Posted" date to at least one minute in the past—this reliably bypasses the issue.
Permanent solution: Report this bug on GitHub using exactly this diagnostic data:
👉 https://github.com/textpattern/textpattern/issues
The diagnostic output you have is perfect for a bug report—
"Database server time offset: 3601 s"is the key piece of evidence. It should actually be 3600, and TXP ought to handle this more robustly.