On 12/25/2016 11:10 AM, G. Schlisio wrote: >> Georg >> >> I don't think there is enough evidence at the moment to say with >> certainty that any change in glibc has introduced the problem, since you >> were using that for a while now without seeing issues. >> >> I'd still be interested in knowing what output the test program gives on >> the affected server. >> >> best wishes to you too. >> >> John > oh, right, the test program. > > me@dukun ~ > ./test2 AAAAAAAA > Not found > me@dukun ~ > echo $? > 1 > > i hope this is the information you looked for? i didnt think about it > too much at the moment… > georg
thanks. That result could cause some headaches. To be consistent with the trace posted at http://termbin.com/ram9 and specifically these lines Dez 23 12:21:44 dukun postfix/local[4396]: deliver_dotforward[3]: local AAAAAAAA recip aaaaa...@dukun.de exten deliver aaaaa...@dukun.de exp_from Dez 23 12:21:44 dukun postfix/local[4396]: warning: error looking up passwd info for AAAAAAAA: Invalid argument the expected result from the test program is a message about invalid argument. You might try to see if postfix is still giving you the same trace as before for the user AAAAAAAA (with the forward_path workaround removed) or if now it is working. That postfix gets one response and the test program a different response on the same server without changing anything else (like nsswitch.conf) or restarting the server etc. is puzzling. I suppose the test you tried in postfix was with that exact same user AAAAAAAA and that is not an anonymization of a longer username. Whatever the case on your server, I think this issue is going to need to be addressed, presumably in archlinux or glibc since I found a reproducable case of your original problem on a stock archlinux install with glibc 2.24: sendmail -bv AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA Dec 25 11:51:48 archlinux postfix/local[2974]: warning: error looking up passwd info for aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa: Invalid argument This is reproducable with the test program (so independently of Postfix). Could you try to get this addressed otherwise more people are going to start finding problems with postfix local delivery on archlinux platform (and potentially with other platforms if the cause is within glibc itself and those platforms start to update to newer releases). I say "if the cause is within glibc" since it may be something that also depends on what nsswitch.conf configurations are being used. In the example below, the getpwnam_r routine returns a correct answer for an inexistent user for strings up to 31 chars. From 32 chars onwards instead of returning not found it retuns EINVAL (invalid argument). /test AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA Not found ./test AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA getpwnam_r: Invalid argument John
