IRC log of #maemo-ssu for Sunday, 2015-03-22

bencohyay :)00:01
*** esoare has quit IRC00:22
*** Pali has quit IRC01:38
*** RedW has quit IRC02:01
*** RedW has joined #maemo-ssu02:06
*** arcean has quit IRC02:17
*** rd_ has joined #maemo-ssu02:29
xesfreemangordon: well done!02:31
*** _rd has quit IRC02:33
*** M4rtinK has quit IRC02:38
*** dafox has quit IRC03:47
*** futpib has quit IRC04:17
*** nox- has quit IRC06:14
*** rd_ has quit IRC09:15
*** arcean has joined #maemo-ssu10:27
*** RzR has quit IRC10:45
*** RzR has joined #maemo-ssu10:46
*** rd_ has joined #maemo-ssu11:02
*** sparetire_ has quit IRC11:13
freemangordonmerlin1991: I think it is relatively safe o include openssl in the next -testing11:37
freemangordon*to11:37
freemangordonif there is something broken, I doubt we'll catch it while the package is in -devel. so far I tested everything I can think about and there seems to be no problem - https, email (imap tls), ssh, supl.11:39
bencohmaybe we should nm all the binaries and grep for affected symbols11:40
bencohmight help catching potential issues11:40
bencohbut the main problem here is things will probably "work" but we might end up with invalid reads/writes at some point11:41
freemangordonbencoh: symbols should be there, it is supposed to be the same ABI11:48
bencohyeah but some structures differ11:49
freemangordonbut you won;t catch that with nm11:49
bencohsome of which might be instantiated from the app and passed to functions11:49
bencohno, but you can grep on the functions which are passed those struct/struct*11:49
freemangordonI doubt there is a transparent structures to be used in applications11:49
freemangordon*are11:49
freemangordonif an application uses such, then I'll call it buggy11:50
freemangordoneven if it malloc's it11:50
bencohthere are structures defined in headers11:50
freemangordonbencoh: but feel free to do ^^^ if you think it is possible11:51
freemangordonso?11:51
bencohthey're not private/opaque :)11:51
bencohso, I dont know11:51
bencohhm, headers look bloated in openssl :D11:52
freemangordon:nod:11:52
keriolet's just move to libressl11:55
keriowe only need to recompile everything11:55
bencoh:/11:56
bencohthere are things we cant rebuild :(11:56
kerio:(11:57
bencohmost most datatype changes reported in http://upstream.rosalinux.ru/versions/openssl.html say "99 symbols affected" (basically every symbole, more or less)12:00
keriowut12:02
bencohthat means nm+grep wont work12:03
freemangordonbencoh: look at the correct version, 0.9.8 that is12:03
kerioyeah i mean12:05
keriowe're going 0.9.8n to 0.9.8zf12:05
freemangordon:nod:12:06
*** Pali has joined #maemo-ssu12:06
bencohfreemangordon: yeah that's what I'm checking :)12:07
bencohlooks like za reverts datatype issues introduced in y12:07
bencohso we're left with issues introduced in p12:08
bencohwhich might be solved just by keeping the old PQ_64BIT behaviour12:09
bencohI wouldnt be surprised if debian did that for us12:09
kerioi like how openssl has shit like "is_probably_safari"12:10
bencoh:))12:11
bencohkerio: then what about PQ_64BIT_IS_BIGNUM (bignum was 20bytes ... 160bits) ? ;p12:11
freemangordonbencoh: could you check the patches?12:12
freemangordonhttps://github.com/community-ssu/openssl/tree/0.9.8zf-1%2Bmaemo1%2B0m5%2B0cssu0/debian/patches12:13
freemangordonhmm, debian has patches up to 0.9.8o :(12:15
freemangordonI guess it is because of the p breaking ABI12:15
bencohis this patch what I think ? https://github.com/community-ssu/openssl/blob/master/debian/patches/12_valgrind.patch12:15
bencohreminds me of an (in)famous debian patch12:16
bencohI think someone had a look at it before, but ...12:16
bencohfreemangordon: hm probably yeah12:16
bencohthey keep backporting changes in squeeze-lts12:17
bencohwhy not use squeeze-lts as upstream for openssl ?12:17
freemangordonhttp://anonscm.debian.org/viewvc/pkg-openssl/openssl/tags/0.9.8o-4squeeze15/debian/patches/12:17
freemangordonbecause the newest patch there is 9 months old12:18
bencohhttps://packages.qa.debian.org/o/openssl/news/20150320T214900Z.html12:18
freemangordonhmm, where did you get that from?12:19
freemangordonI mean - where is the source repo?12:19
bencohdunno12:19
bencohI didnt get it - yet12:19
bencohand ... google cant find the packages12:20
bencohhttp://ftp.de.debian.org/debian/pool/main/o/openssl/openssl_0.9.8o-4squeeze20.debian.tar.gz12:21
bencohgoogle hasnt indexed it12:21
bencoh(yet)12:21
freemangordonwe'd rather fix the ABI breakage (if any)12:23
bencohthe valgrind.patch is still in debian, so I guess (hope ? :D) it's not the infamous 2008ish debianssl bg12:23
bencohbug12:24
bencohfreemangordon: there our best bet is probably defining PQ_64BIT_IS_BIGNUM to 1 and PQ_64BIT_IS_INTEGER to 012:24
bencoh(and maybe the other two PQ_64BIT constants)12:24
freemangordonbencoh: look at pqueue/pq_compat.h12:27
bencohyeah12:29
freemangordonand it seems defined in SB12:29
freemangordon~/workspace/cssu/openssl/git/openssl-0.9.8zf] > grep -r THIRTY *12:29
bencohoh ?12:29
freemangordon...12:29
freemangordoncrypto/opensslconf.h:#define THIRTY_TWO_BIT12:29
bencoh:D12:29
bencohnice, we might be safe then12:30
freemangordonI'll put some garbage in that header, to make sure this is really defained12:30
bencohhm we need to make sure it's used everywhere12:31
freemangordon../../include/openssl/pq_compat.h:113:2: error: #error "# if (defined(THIRTY_TWO_BIT) && !defined(BN_LLONG)) || defined(SIXTEEN_BIT) || defined(EIGHT_BIT) *IS NOT* defined, we are busted :("12:35
freemangordonmake[3]: *** [pqueue.o] Error 112:35
bencohokay, that's the only place PQ_64BIT is defined, so we should be safe12:35
bencohoh ?12:35
bencohhmm right, we're 32b but we have long long support on arm12:36
freemangordonyeah12:38
freemangordonthat one should be fixed. I guess12:38
freemangordonhowever, I gtg, will look at it later12:38
bencohthat means we give up on long long optim ... well, if those didnt exist before 0.9.8p, we wont miss anything12:39
freemangordonwell, we can try to wrap12:41
freemangordonthose are just a couple of functions anyway12:42
bencohyeah12:43
bencohfunny thing is ... I cant find those changes in openssl git (git log between tags) :/12:43
bencohdunno what they did there12:43
freemangordoncheck on github, they have "file history"12:44
freemangordonhttps://github.com/openssl/openssl12:44
bencoh?12:44
freemangordonhttps://github.com/openssl/openssl/commits/OpenSSL_0_9_8zf/crypto/pqueue/pqueue.h12:45
freemangordonhttps://github.com/openssl/openssl/commit/e2ac4732cd4f0efc3dff3261d19aed0ced2c4dde12:45
freemangordonhmm, wait, isn;t that pq shit internal?12:46
bencohno, it's reflected in public structures12:46
bencohdtls1_/ssl3_ stuff12:47
freemangordonhmm, yeah12:47
freemangordonhttps://github.com/openssl/openssl/commit/e2ac4732cd4f0efc3dff3261d19aed0ced2c4dde#diff-bfa5a0561be6e4126229a7fd57bc24e0R29912:47
bencohssl3_record_st has seq_num for instance12:47
bencohright ;)12:47
freemangordonok, we can wrap it12:48
freemangordonor maybe not12:48
bencoh:]12:48
freemangordonwait... do we really need to fix that?12:49
bencohdunno12:49
freemangordonthe new structure is smaller than the old one12:49
freemangordonmerlin1991: Pali: ^^^12:50
*** amiconn has quit IRC12:50
freemangordonmaybe someone should ask on #openssl12:52
freemangordonor whateer the channel is12:52
*** amiconn has joined #maemo-ssu12:52
* freemangordon is afk12:53
PaliI really do not know :-(12:53
freemangordonPali: noone knows, that is why we investigate :)12:53
freemangordonbencoh: could you check what sizeof(those 2 structs) gives with old and with new openssl?12:58
bencohI'm checking diffs between (current cssu) openssl-0.9.8n and zf, and ...12:58
bencohdiff -w -u crypto/pqueue/pq_compat.h ../openssl-cssu/openssl-0.9.8zf/crypto/pqueue/pq_compat.h12:58
bencohreturns nothing12:58
bencohI feel like I missed something at some point12:59
bencohor upstream.rosalinux.ru says shit12:59
freemangordonkeep it mind that it might say shit, becuse of the missing arch specific stuff13:01
bencohtrue13:01
bencohit might just be that13:01
freemangordonthat is why the real test should be comparing the sizeof structs, just to be sure13:01
freemangordonbut I don;t have time now13:02
freemangordonbuy, bbl13:02
bencohshould I just compile a oneliner against the headers ?13:02
*** M4rtinK has joined #maemo-ssu13:28
*** futpib has joined #maemo-ssu13:41
bencohI should have disabled the checks in debian/rules ... it's taking ages13:55
*** trx has joined #maemo-ssu14:02
*** rd_ has quit IRC14:08
bencohfreemangordon: http://pastebin.notk.org/pastebin.php?show=m6cc01d7d14:10
bencohlooks good14:10
*** rd_ has joined #maemo-ssu14:17
Paliwhy is git read-tree so sloooow?14:34
Paliprobably it removes all inodes information from .git/index...14:35
*** rd_ has quit IRC14:58
*** rd_ has joined #maemo-ssu15:00
bencohhttp://pastebin.notk.org/pastebin.php?show=m597f644e15:03
Palionly two structs?15:22
Palithis is not too much15:23
*** arcean has quit IRC15:28
*** arcean has joined #maemo-ssu15:38
*** arcean has quit IRC15:51
bencohsome structs might be missing in this test, but at least it covers the pq64 thing, and a few others16:08
freemangordonbencoh: yes, looks good17:07
freemangordonexcept that x509_cinf_st thingie17:08
bencohyeah, dunno where that comes from17:08
freemangordonbencoh: does headers differ?17:12
freemangordonit is crypto/x509/x509.h17:12
bencohhmm right, looks like they added ASN1_ENCODING enc17:15
bencohcouldnt see it in git log :)17:15
freemangordonbencoh: https://github.com/openssl/openssl/commit/6cb5746b65a53207bc794121a0c78603bff8ea3017:16
bencohand hm_fragment_st has unsigned char *reassembly17:16
freemangordonbencoh: anyway X509_CINF structure should not be allocated with malloc, there is X509_CINF *X509_CINF_new(); function17:18
bencohhmm, so we should be fine17:18
bencohhm_fragment doesnt seem to be used in any public function (?!)17:19
freemangordonI don't see it used at all, weird17:22
freemangordonoh17:22
freemangordonit is hw_fragment17:22
freemangordonyep, seems like internal struct17:23
freemangordonbencoh: seems the ABI is the same17:23
bencohhm_* not hw_*17:24
freemangordonyeah, a typo17:24
bencohbut ... yeah, everything's fine in the end17:24
freemangordon:)17:24
bencohI guess I just worried too much :)17:24
freemangordonno, it is ok, better be on the safe side with such a critical library17:25
*** futpib has quit IRC18:06
*** rd_ has quit IRC18:29
*** rd_ has joined #maemo-ssu18:29
*** xes_ has joined #maemo-ssu18:31
*** xes has quit IRC18:33
*** LauRoman has joined #maemo-ssu18:46
keriofreemangordon: do i yolo upgrade19:13
kerioare there some reported breakages19:14
freemangordonkerio: nothing broken so far19:15
kerioalright i found a packaging problem already19:15
kerioopenssl doesn't depend on libssl0.9.8 >= 0.9.8zf19:15
freemangordonI even managed to connect to FB with QWebView19:15
freemangordonsure, why it should depend on that particular varion?19:15
freemangordon*version19:16
keriobecause... it's openssl 0.9.8zf19:16
freemangordonbut, ABI is 0.9.819:16
keriok19:16
freemangordonnot 0.9.8zf19:16
keriofreemangordon: there's a new busybox-power apparently19:17
freemangordonkerio: I have only 2 hands, still19:18
freemangordonand 24hours per day19:18
freemangordondeffinitely, openssl is of higher priority19:18
keriofreemangordon: http://www.blogcdn.com/www.engadget.com/media/2008/05/5-29-08-carpet-keyboard.jpg19:18
freemangordonthose are legs, not hands19:19
keriomh, who uploaded libc6 +cssu3?19:21
freemangordonmerlin1991, afaik19:22
keriowhy was a non-thumb libc6 in your repo19:22
merlin1991kerio: +cssu3 is in -devel not in -thumb19:23
kerio+cssu2 is in -thumb19:23
*** LauRoman has quit IRC19:58
*** rd_ has quit IRC20:03
freemangordonkerio: because thumb-compiled seemed to bring ptoblems20:09
kerio;-;20:09
freemangordonso I increased the version, ARM-compiled it and put it in the repo20:09
*** LauRoman has joined #maemo-ssu20:11
keriok20:13
*** sparetire_ has joined #maemo-ssu20:32
*** rd_ has joined #maemo-ssu20:39
*** arcean has joined #maemo-ssu20:41
freemangordonPali: did you backout linux-n900 on github?20:43
freemangordon*backup20:43
*** M4rtinK has quit IRC20:45
*** futpib has joined #maemo-ssu20:47
Palifreemangordon: week ago... https://github.com/pali/linux-n90020:53
freemangordonPali: ok20:55
freemangordonkerio: http://maemo.merlin1991.at/cssu/community-thumb/pool/free/b/busybox-power/20:56
kerioyey20:56
*** LauRoman has quit IRC21:01
*** freemangordon has quit IRC21:10
*** freemangordon has joined #maemo-ssu21:11
*** M4rtinK has joined #maemo-ssu21:25
*** M4rtinK has quit IRC21:35
*** M4rtinK has joined #maemo-ssu21:38
*** M4rtinK has quit IRC22:10
*** arcean has quit IRC22:12
*** arcean has joined #maemo-ssu22:14
*** M4rtinK has joined #maemo-ssu22:23
*** M4rtinK has quit IRC22:27
*** futpib has quit IRC22:40
*** futpib has joined #maemo-ssu22:40
*** futpib has quit IRC22:45
*** futpib has joined #maemo-ssu23:04
*** futpib has quit IRC23:09
*** futpib_ has joined #maemo-ssu23:10
*** futpib_ has quit IRC23:15
*** dafox has joined #maemo-ssu23:21

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!