28.04.2011 18:22, Wietse Venema пишет: > Michael Tokarev: >> 28.04.2011 15:08, Wietse Venema wrote: >>> Michael Tokarev: >>>> postfix/cleanup: warning: milter8_message: vstream_fseek >>>> incoming/4BE085028D: File too large >>> >>> Why is this reported as a "450 4.3.0" error? This should >>> be a permanent error instead. >> >> No it shouldn't. Ok, _this_ very condition - EFBIG - >> should be, indeed, but not other possible I/O errors. > > Indeed. There is code that correctly flags EFBIG as a permanent > error in cleanup_milter.c, cleanup_final.c, and cleanup_out.c. > > This is easy enough to patch into the Milter client but it seems > wrong to have EFBIG checks in so many different places, because > it is easy to overlook an EFBIG check. > > As I researched the problem I found that milter_header_checks also > missed this check. All this is supposedly fixed with the attachedd > patch.
The patch appears to fix the issue. Actually I added the part in milter8.c into my postfix even before sending the initial report, to let the message to bounce correctly - exactly the same hunk as you added there. I don't see how to trigger this issue in milte_header_checks using milters I have handy, but indeed, it's of the same nature and looks troublesome too. It's curious how this issue has been here for so long and no one noticed it so far :) Thank you! /mjt
