Bug#784115: inn: incoming feed is garbled
Hi Julien, On Sat, May 30, 2015 at 12:26:58PM +0200, Julien ÉLIE wrote: > So maybe the issue comes from Perl... Is it possible to disable the > Perl filters in inn.conf to see if it then works OK? I don't know where/how to disable the filters in the configuration but I disabled them immediately after starting the daemon with $ ctlinnd perl n This did not make any change, I still get those errors. > Or a large-file support issue? > Is the problem only with the amd64 package? > [...] Sorry, but I can't try this here :-( Stefan -- Für den anspruchsvollen Herren - hüpfen mit Stefan! http://www.sloganizer.de/ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#784115: inn: incoming feed is garbled
Hi Stefan, Good: inn_1.7.2q-41_amd64.deb Fail: inn_1.7.2q-41+b1_amd64.deb I don't know anything about the differences of these two versions. However, 41 has been running without a single error for a couple of hours now. When I switched to 41+b1 again, errors came within seconds. The only change between -41 and -41b1 is a rebuild against perl 5.18, isn't it? So maybe the issue comes from Perl... Is it possible to disable the Perl filters in inn.conf to see if it then works OK? Or a large-file support issue? (I do not know if LFS also exists in INN 1.7.2 but one should note that we had issues with INN 2.x against latest versions of Perl.) Is the problem only with the amd64 package? Fixing the pointer-cast-size-mismatch warnings may be helpful if that is the case: https://qa.debian.org/bls/packages/i/inn.html We can see the warnings in: https://buildd.debian.org/status/fetch.php?pkg=inn&arch=amd64&ver=1%3a1.7.2q-44%2bb2&stamp=1412186175 I notably see that one concerns Perl in the CHECK command (used for streaming!): nc.c: In function 'NCcheck': nc.c:1592:24: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast] else if ((perlrc = (char *)HandleMessageID(p)) != NULL) { -- Julien ÉLIE « Ce n'est pas en tournant le dos aux choses qu'on leur fait face. » (Pierre Dac) -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#784115: inn: incoming feed is garbled
On Thu, May 28, 2015 at 06:26:36PM +0200, Stefan Froehlich wrote: > On Thu, May 28, 2015 at 03:58:51PM +0200, Marco d'Itri wrote: > > > Good: inn_1.7.2q-41_amd64.deb > > > Fail: inn_1.7.2q-41+b1_amd64.deb > I don't know anything about the differences of these two versions. > However, 41 has been running without a single error for a couple of > hours now. When I switched to 41+b1 again, errors came within seconds. To justify this I did straces on these two versions. Have a look at the two calls to writev() in the respective attachment. strace-inn41.txt reads 2818 bytes and writes 2750. The difference of 68 bytes comes from the "TAKETHIS" string at the beginning and the conversion of \r\n to \n, but otherwise the posting is stored without modification. strace-inn41+b1.txt reads 2641 bytes and writes 2502. In addition to the expected difference mentioned above there is a part of the signature missing: | read(17, "[...] Beginning of Wisdom \" | =\r\nhttp://www.zugschlus.de/\r\nNordisch by Nature | Lt. Worf, TNG \"Rightful Heir\" | Fon: *49 621 =\r\n72739834\r\n.\r\n", 4095) = 2641 | [...] | writev(18, [{"Path: ", 6}, {"hetzner2.epaxios.com!", 21}, {"87.106.167.149", 14}, {"!", 1}, {"[...] Beginning of Wisdom \" | =\nhttp: *49 621 =\n72739834\n", 1439}], 8) = 2502 And as always, a part of the missing text (but neither the exact beginning nor the exact end of the gap) appears as a "bad command" in the error message: | sendto(3, "<61>May 30 09:11:50 innd: pasture.szaf.org:17 bad_command l Heir\" | Fon: *49 621 =", 82, MSG_NOSIGNAL, NULL, 0) = 82 So there *must* be a significant difference between these two versions. In fact I did not find a single article without errors in 41+b1. Could it be helpful if I provide a dozen of them in order to find some common pattern? Stefan read(17, "TAKETHIS \r\nPath: news.szaf.org!weretis.net!feeder1.news.weretis.net!news1.tnib.de!feed.news.tnib.de!news.tnib.de!.POSTED!not-for-mail\r\nFrom: Marc Haber \r\nNewsgroups: de.rec.tv.technik\r\nSubject: Re: Fehlersuche in Sat-Anlage mit Multischalter\r\nDate: Fri, 29 May 2015 20:40:24 +0200\r\nOrganization: private site, see http://www.zugschlus.de/ for details\r\nLines: 50\r\nMessage-ID: \r\nReferences: \r\nMime-Version: 1.0\r\nContent-Type: text/plain; charset=utf-8\r\nContent-Transfer-Encoding: quoted-printable\r\nX-Trace: news1.tnib.de 1432924824 3868 37.49.103.85 (29 May 2015 18:40:24 GMT)\r\nX-Complaints-To: ab...@tnib.de\r\nX-Newsreader: Forte Agent 6.00/32.1186\r\nXref: news.szaf.org de.rec.tv.technik:40311\r\n\r\nMarc Haber wrote:\r\n>ich habe hier ein kleines Problem mit der Sat-Anlage im (vermieteten)\r\n>Haus meiner Frau. Ich vermute, der Multischalter ist kaputt.\r\n\r\nHeute war der Elektriker endlich da. Er hat - ohne sich vorher das\r\n=46ehlerbild im Haus anzuschauen - auf dem Dach die Sch=C3=BCssel neu\r\njustiert und alle Stecker des LNB =C3=BCberpr=C3=BCft. Als sich dann im =\r\nHaus das\r\ngleiche Fehlerbild weiterhin zeigt, konstatierte er einen defekten LNB\r\nund k=C3=BCndigte direkt an, dass das nur ein =C3=9Cberspannungsschaden =\r\nsein\r\nk=C3=B6nnte und dass der Tausch des LNB nicht auf Gew=C3=A4hrleistung =\r\nginge.\r\n\r\nBei der dann folgenden hochnotpeinlichen Befragung verstrickte er sich\r\nderart in Widerspr=C3=BCche, dass ich kurz in die Tischkante bei=C3=9Fen =\r\nund\r\ndanach ein wenig laut werden musste. Der Elektriker kehrte daraufhin\r\nzu Multischalter zur=C3=BCck, erkannte nach wenigen weiteren Minuten der\r\nMessung die zwei vertauschten Downlink-Kabel und schwupps waren alle\r\nvier Ebenen dort, wo sie hingeh=C3=B6ren.\r\n\r\nIch freue mich schon auf die Diskussion, wenn die Rechnung kommt, was\r\nzweifelsfrei passieren wird.\r\n\r\nSo ganz erkl=C3=A4ren kann ich mir meine Me=C3=9Fergebnisse in diesem =\r\nThread mit\r\ndiesem Fehler aber nicht. Kann das jemand erkl=C3=A4ren?\r\n\r\nInteressant finde ich =C3=BCbrigens, dass sein Me=C3=9Fger=C3=A4t[1] =\r\nsofort angezeigt\r\nhat, welche Ebene auf diesem Downlink-Kabel liegt. Gibt es vielleicht\r\nim Spektrom doch einen Marker, den ein Sat-Me=C3=9Fger=C3=A4t direkt =\r\nanzeigen\r\nkann?\r\n\r\nGr=C3=BC=C3=9Fe\r\nMarc\r\n\r\n[1] er hatte drei dabei, weil man \"auf dem modernen Zeug\" nicht mehr\r\nalles einstellen kann\r\n--=20\r\n-- !! No courtesy copies, please !! =\r\n-\r\nMarc Haber | \" Questions are the | Mailadresse im =\r\nHeader\r\nMannheim, Germany | Beginning of Wisdom \" | =\r\nhttp://www.zugschlus.de/\r\nNordisch by Nature | Lt. Worf, TNG \"Rightful Heir\" | Fon: *49 621 =\r\n72739834\r\n.\r\n", 8192) = 2818 stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=2309, ...}) = 0 lseek(7, 471040, SEEK_SET) = 471040 read(7, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0", 32) = 32 read(7, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\
Bug#784115: inn: incoming feed is garbled
On Thu, May 28, 2015 at 03:58:51PM +0200, Marco d'Itri wrote: > > Good: inn_1.7.2q-41_amd64.deb > > Fail: inn_1.7.2q-41+b1_amd64.deb > I do not think that this is right, because 41+b1 is just a binNMU which > is supposed to be in every way identical to 41. I don't know anything about the differences of these two versions. However, 41 has been running without a single error for a couple of hours now. When I switched to 41+b1 again, errors came within seconds. Sorry... Bye, Stefan -- Stefan - die vornehmste Ueberraschung der Poesie! http://www.sloganizer.de/ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#784115: inn: incoming feed is garbled
On May 28, Stefan Froehlich wrote: > Good: inn_1.7.2q-41_amd64.deb > Fail: inn_1.7.2q-41+b1_amd64.deb I do not think that this is right, because 41+b1 is just a binNMU which is supposed to be in every way identical to 41. If the problem is really between 41 and 42 then I am screwed, because it is when the conversion to git happened. :-( -- ciao, Marco pgpU2e93kGBvP.pgp Description: PGP signature
Bug#784115: inn: incoming feed is garbled
On Sun, May 10, 2015 at 11:36:59PM +0200, Marco d'Itri wrote: > What would really help is if you could test some of the older packages > to find out which one introduced this regression: > http://snapshot.debian.org/package/inn/ Good: inn_1.7.2q-41_amd64.deb Fail: inn_1.7.2q-41+b1_amd64.deb (There is no inn_1.7.2q-42_amd64.deb available, inn_1.7.2q-43_amd64.deb and above all fail) Bye, Stefan -- Geht nicht!? Das gibt's nicht, jedenfalls nicht bei uns: Stefan. http://www.sloganizer.de/ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#784115: inn: incoming feed is garbled
On May 10, Marco d'Itri wrote: > What would really help is if you could test some of the older packages > to find out which one introduced this regression: > http://snapshot.debian.org/package/inn/ BTW, you can just replace on the fly the innd binary: there is no need to downgrade/reinstall the package. -- ciao, Marco pgpO7bkBjMPxw.pgp Description: PGP signature
Bug#784115: inn: incoming feed is garbled
On May 04, Stefan Froehlich wrote: > A little bit unexpected but luckily: the problems stopped immediately > when switching to non-streaming. Not so unexpected, they are very different code paths. > > Another interesting test would be to manually start innd without using > > systemd's socket activation protocol. > This did not make any difference however. There are not many other changes, so this bu is very non-obvious. :-( > The workaround with on-streaming is sufficent for me. However, if you > want to dig into this to fix the acutal bug, I can do/change/try > whatever you want to provide you with further details. What would really help is if you could test some of the older packages to find out which one introduced this regression: http://snapshot.debian.org/package/inn/ -- ciao, Marco pgpPk31Yvn8Nn.pgp Description: PGP signature
Bug#784115: inn: incoming feed is garbled
On Mon, May 04, 2015 at 01:58:54AM +0200, Marco d'Itri wrote: > Can you check if a non-streaming feed (IHAVE only, i.e. "streaming: > false" in innfeed.conf) triggers the bug too? A little bit unexpected but luckily: the problems stopped immediately when switching to non-streaming. > Another interesting test would be to manually start innd without using > systemd's socket activation protocol. This did not make any difference however. The workaround with on-streaming is sufficent for me. However, if you want to dig into this to fix the acutal bug, I can do/change/try whatever you want to provide you with further details. Stefan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#784115: inn: incoming feed is garbled
On May 03, news wrote: > Unfortunately I have no idea WHAT would cause this and thus as > well have no suggestion how to fix it. The error is easily Me neither, I do not see any obvious changes since oldstable (and I do not use NNTP with INN 1.x): http://anonscm.debian.org/cgit/users/md/inn.git/log/ Can you check if a non-streaming feed (IHAVE only, i.e. "streaming: false" in innfeed.conf) triggers the bug too? Another interesting test would be to manually start innd without using systemd's socket activation protocol. -- ciao, Marco pgptrci_xPVwB.pgp Description: PGP signature
Bug#784115: inn: incoming feed is garbled
Package: inn Version: 1:1.7.2q-44+b2 Severity: important Dear Maintainer, *** Reporter, please consider answering these questions, where appropriate *** Inn was installed to a fresh Jessie system, minor changes have been done to the configuration (see attachments) and a single feed was added. As soon as the feed began to transmit, strange error messages showed up, claiming "bad_command"s having been sent. I did several straces of tin trying to understand what happens. One example is shown in the attachment inn.strace. It corresponds to the log entries | May 3 10:23:01 hetzner2 innd: pasture.szaf.org connected 17 streaming allowed | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 NCmode "mode stream" received | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_messageid @mid.individual.net> | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command net> | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command 3e | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command t>3epF3jf6ual.ividuaividuaividuaividual.net> | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command 63 Zeichen. | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command Viele s | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command ., | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command Johanne . ., Johanne ., ohaJoha Joha Joh Johannes s . . es s . Joh... | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command ., | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command Johanne ., s sJohansJohanneannesnnes s . . nes s . . es s . Joh... | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command ., sJohannes s | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command . | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 bad_command s s | May 3 10:23:01 hetzner2 innd: pasture.szaf.org:17 closed seconds 0 accepted 1 refused 1 rejected 0 Please note especially line 48 (with TAKETHIS) and line 51. In line 49 an article is received over the network, in line 51 the same article is written to /var/spool/news/de/sci/mathematik/15 in the local file system. Note the fact that 2056 bytes are read (with 42 bytes overhead) but only 1919 bytes are written. If you compare the two payloads of read() and write(), a couple of bytes are missing in the locally stored version of the article. (And the following "500" error messages cite text which is close to that garbled part of the posting. The very same thing (with similar, but not identical amounts of data missing) happens with the major part of articles transferred. This seems to me as a clear case of data loss - and if the server had an outgoing feed, the wrong versions of the articles would be spread into the world. Unfortunately I have no idea WHAT would cause this and thus as well have no suggestion how to fix it. The error is easily reproducible however (in fact, I get reports every couple of minuates), so if additional information is required, feel free to ask. Stefan *** End of the template - remove these template lines *** -- System Information: Debian Release: 8.0 APT prefers stable-updates APT policy: (500, 'stable-updates'), (500, 'stable') Architecture: amd64 (x86_64) Kernel: Linux 3.16.0-4-amd64 (SMP w/8 CPU cores) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Init: systemd (via /run/systemd/system) Versions of packages inn depends on: ii bsd-mailx [mailx]8.1.2-0.20141216cvs-2 ii cron 3.0pl1-127 ii init-system-helpers 1.22 ii libc62.19-18 ii libperl5.20 5.20.2-3 ii libsystemd0 215-17 ii perl 5.20.2-3 ii perl-base [perlapi-5.20.1] 5.20.2-3 ii sendmail-bin [mail-transport-agent] 8.14.4-8 ii time 1.7-25 inn recommends no packages. Versions of packages inn suggests: ii gnupg 1.4.18-7 -- Configuration Files: /etc/news/hosts.nntp changed: localhost/t: pasture.szaf.org: /etc/news/inn.conf changed: organization: Reusenet server: localhost /etc/news/newsfeeds changed: ME:*,!control,!junk,!local.*/!local:: overview:*,!control.cancel:Tc,WO:/usr/lib/news/bin/overchan reusenetfilesystemgateway:*:Tf,Wf,Freusenetfilesystemgateway: -- no debconf information Process 6937 attached select(14, [3 11 13], [], NULL, {299808, 922648}) = 1 (in [3], left {299804, 958833}) accept(3, {sa_family=AF_INET, sin_port=htons(55887), sin_addr=inet_addr("87.106.167.149")}, [16]) = 17 open("/etc/protocols", O_RDONLY|O_CLOEXEC) = 18 fstat(18, {st_mode=S_IFREG|0644, st_size=2932, ...}) = 0 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa3cd2ca000 read(18, "# Internet (IP) protocols\n#\n# Updated from http://www.iana.org/assignments/protocol-numbers and