Hello there After upgrading a Debian server from 7.8 (with Apache 2.2) to 8.2 (with Apache 2.4) I started seeing a very curious problem, of error_log lines being split randomly: [Mon Nov 16 13:29:54.388526 2015] [cgi:error] [pid 9122] [client xx.xx.xx.xx:1865] AH01215: (Fred, Test) Warning: foo.cgi/Foo.pm::foo():115 "Argument "test" isn't numeric i, referer: http://... [Mon Nov 16 13:29:54.390770 2015] [cgi:error] [pid 9122] [client xx.xx.xx.xx:1865] AH01215: n numeric eq (==)" [GET http://...], referer: http://... In this example, the word "in" (from "... isn't numeric in numeric eq ... ") is split across two lines. The behaviour appears to be entirely random in terms of which lines it splits and where in the line the split occurs. The only thing I've noticed is that it appears to occur when multiple lines of text are written to STDERR in a row by the same CGI script. It gives the appearance like the STDERR stream is being received by Apache in blocks of (say) 1024 bytes, with a line break being added at the end of each block of text regardless of whether more data is waiting to be read. The varying length of lines written by the CGI script to STDERR would of course give such a bug a random appearance. Does this ring true with anyone? Has anything changed in terms of how Apache receives the STDERR stream from basic CGI scripts? This is just basic CGI: ScriptAlias and Perl scripts (no FastCGI or anything). The Apache version is Apache/2.4.10 (Debian), in Debian 8.2 64-bit. Regards Daniel. --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
|