https://bz.apache.org/bugzilla/show_bug.cgi?id=69264
--- Comment #8 from Guido Jäkel <g.jae...@dnb.de> --- I've thought about Ruedigers suggestion to check if apr_file_write_full() will report bytes written as commissioned. I'm currently using apr-1.7.4. Staring at the goats, aeh code of apr_file_write_full() (https://github.com/apache/apr/blob/b4197909d26042ece777727cc6152deeea75a3ab/file_io/unix/fullrw.c#L42) this uses a vanilla copy loop. I can't see any possibility either to return successful; and then the calculated bytes_written MUST be the number of nbytes instructed to transfer. Or to return unsuccessful with a different value, which will be reported by the suggested patch. But the end of the string is received and this CAN'T happen, if the loop exit before reaching it's normal end. Therefore, it MUST have instructed to transfer any slice before the end, including the start. In other words, I don't see any logic to skip over some leading parts of the string. The copy loop uses apr_file_write() (https://github.com/apache/apr/blob/b4197909d26042ece777727cc6152deeea75a3ab/file_io/unix/readwrite.c#L147) to do thew work. "Loosing" something to transfer while reporting it to be done MUST be happen here. Therefore, IMHO it must be a bug or thread race condition inside this much more complex code. What do you think about? -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: bugs-unsubscr...@httpd.apache.org For additional commands, e-mail: bugs-h...@httpd.apache.org
