On Sun, 12 Feb 2017, Brian May wrote: > >> https://github.com/perl5-dbi/DBD-mysql/issues/78 > > Please read the latest message on the above report...
Jakub, This issue is a nasty problem between amavisd-new and DBD-mysql. It is yet not clear which program is at fault, or if it is an ABI break gone bad, etc. The upstream bug report is at: https://github.com/perl5-dbi/DBD-mysql/issues/78 The amavisd-new Debian maintainers and DBD upstream are having trouble to reproduce the issue, although we do have breakage reports also from in Fedora, Gentoo and FreeBSD. There is cleary some factor we're missing. Upstream DBD asks (in the issue linked above): "So I see only option now, enabling DBI tracing in that problematic amavis code and sending stderr trace output for later inspection. Tracing can be enabled by DBI->trace(2); and disabled by DBI->trace(0);" Would you be willing to reproduce the issue you reported with the changes requested by upstream? AFAIK, you just need to add: $conn_h->trace(2); around line 20211 in /usr/sbin/amavisd-new (I think right after or right before the $conn_h->begin_work_nontransaction; # (re)connect if not connected line would work best. It is probably best to disable the tracing at the end of the function, by adding: $conn_h->trace(0); right before the line reading "section_time('lookup_sql');" around line number 20262. If that doesn't work, try using DBI->trace(2); and DBI->trace(0); instead. Since the trace information goes to stderr, it would probably be best to debug amavis by running it in "foreground" mode, and diverting stderr to a file... Thanks! -- Henrique Holschuh
