Switch to DuckDuckGo Search
   February 19, 2020  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | >

Toggle Join/Part | bottom
[00:07:57] *** Kurlon <Kurlon!~Kurlon@cpe-67-253-136-97.rochester.res.rr.com> has joined #illumos
[00:11:15] *** andy_js <andy_js!~andy@> has quit IRC (Quit: andy_js)
[00:24:33] *** ptribble <ptribble!~ptribble@cpc92716-cmbg20-2-0-cust138.5-4.cable.virginm.net> has quit IRC (Quit: Leaving)
[00:31:42] <LeftWing> richlowe: Oof
[00:34:45] <alanc> https://pubs.opengroup.org/onlinepubs/9699919799/functions/getdate.html - how fun
[00:40:44] <LeftWing> It is not a great interface
[00:42:59] *** igork <igork!~igork@> has quit IRC (Ping timeout: 260 seconds)
[00:45:53] *** igork <igork!~igork@> has joined #illumos
[02:36:49] *** neirac <neirac!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Read error: Connection reset by peer)
[02:37:13] *** neirac <neirac!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[02:58:42] *** tg2 <tg2!~tg2@> has quit IRC (Ping timeout: 265 seconds)
[03:01:12] *** PMT <PMT!~PMT@pool-100-33-69-155.nycmny.fios.verizon.net> has quit IRC (Quit: leaving)
[03:08:10] *** PMT <PMT!~PMT@pool-100-33-69-155.nycmny.fios.verizon.net> has joined #illumos
[03:25:07] *** tg2 <tg2!~tg2@> has joined #illumos
[03:34:26] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has quit IRC (Quit: jcea)
[03:34:26] *** nde <nde!uid414739@gateway/web/irccloud.com/x-ctfzyukdpufuciro> has quit IRC (Quit: Connection closed for inactivity)
[03:38:55] *** tg2 <tg2!~tg2@> has quit IRC (Ping timeout: 260 seconds)
[03:57:39] *** tg2 <tg2!~tg2@> has joined #illumos
[04:07:46] <glasspelican> im working on 11902 and im not sure if i should add myself to the copyright for the file for a simple spelling fix
[04:28:30] <idodeclare> Technically: CDDL 3.3. Required Notices. You must include a notice in each of Your Modifications that identifies You as the Contributor of the Modification....
[04:44:46] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[04:47:59] *** neirac <neirac!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 260 seconds)
[04:58:50] <glasspelican> Ok, I'll make that change and try to get it sent off to the developer mailing list tomorrow
[05:30:55] *** tg2 <tg2!~tg2@> has quit IRC (Ping timeout: 260 seconds)
[05:54:54] *** tg2 <tg2!~tg2@> has joined #illumos
[06:02:44] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[06:06:22] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 260 seconds)
[06:30:15] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has quit IRC (Remote host closed the connection)
[06:52:34] *** _alhazred <_alhazred!~Alex@> has joined #illumos
[06:55:31] *** _alhazred <_alhazred!~Alex@> has quit IRC (Client Quit)
[06:59:19] <tsoome> glasspelican in practice the modifications are recorded in form of git commit and the change can be tracked there. If you want to list yourself with copyright line is up to you:)
[07:07:41] *** morgib <morgib!uid342182@gateway/web/irccloud.com/x-ancvtrwunqfjclcz> has joined #illumos
[07:10:54] *** _alhazred <_alhazred!~Alex@> has joined #illumos
[07:32:49] *** idodeclare <idodeclare!~textual@cpe-76-185-177-63.satx.res.rr.com> has quit IRC (Quit: My MacBook has gone to sleep. ZZZzzz…)
[07:58:54] *** tsoome <tsoome!~tsoome@148-52-235-80.sta.estpak.ee> has quit IRC (Quit: This computer has gone to sleep)
[07:59:39] *** idodeclare <idodeclare!~textual@cpe-76-185-177-63.satx.res.rr.com> has joined #illumos
[08:00:11] *** glasspelican <glasspelican!~quassel@2607:5300:201:3100::325> has quit IRC (Quit: http://quassel-irc.org - Chat comfortably. Anywhere.)
[08:01:14] *** glasspelican <glasspelican!~quassel@2607:5300:201:3100::325> has joined #illumos
[08:16:14] *** ptribble <ptribble!~ptribble@cpc92716-cmbg20-2-0-cust138.5-4.cable.virginm.net> has joined #illumos
[08:21:39] *** neuroserve <neuroserve!~toens@> has joined #illumos
[08:23:08] <Agnar> moin
[08:23:11] *** BOKALDO <BOKALDO!~BOKALDO@> has joined #illumos
[08:23:27] <neuroserve> re
[08:29:58] *** wiedi <wiedi!~wiedi@ip5b4096a6.dynamic.kabel-deutschland.de> has quit IRC (Quit: ^C)
[08:41:39] *** tsoome <tsoome!~tsoome@148-52-235-80.sta.estpak.ee> has joined #illumos
[08:52:17] *** kayront- <kayront-!~kayront@zbase.xen.prgmr.com> has left #illumos ("Leaving")
[08:54:40] *** jimklimov <jimklimov!~jimklimov@ip-86-49-254-26.net.upcbroadband.cz> has joined #illumos
[09:06:54] *** _alhazred <_alhazred!~Alex@> has quit IRC ()
[09:12:29] *** basvdlei <basvdlei!~basvdlei@2001:980:a4c3:1:b88b:1590:d6bf:fbee> has quit IRC (Quit: basvdlei)
[09:13:12] *** wiedi <wiedi!~wiedi@> has joined #illumos
[09:16:23] *** hemi770 <hemi770!~hemi666@unaffiliated/hemi770> has quit IRC (Quit: The life of a repo man is intense)
[09:18:18] <wilbury> moin
[09:22:11] <leoric> Does someone else want to look at http://buildzone.oi-build.r61.net/webrev-12316/ ?
[09:35:16] *** morgib <morgib!uid342182@gateway/web/irccloud.com/x-ancvtrwunqfjclcz> has quit IRC (Ping timeout: 248 seconds)
[09:35:17] *** scoobybejesus <scoobybejesus!sid271506@gateway/web/irccloud.com/x-lqmqlqzanwzjngpt> has quit IRC (Ping timeout: 248 seconds)
[09:37:30] *** scoobybejesus <scoobybejesus!sid271506@gateway/web/irccloud.com/x-edhvdfyhqytihgah> has joined #illumos
[09:38:36] *** morgib <morgib!uid342182@gateway/web/irccloud.com/x-ydklzcflmwqamloe> has joined #illumos
[09:57:10] *** morgib <morgib!uid342182@gateway/web/irccloud.com/x-ydklzcflmwqamloe> has quit IRC (Quit: Connection closed for inactivity)
[10:00:02] <toasterson> @freenode_neirac:matrix.wegmueller.it: gladly. I am interested to get Other tools like nomad/ansible/terraform working aswell on it, so users can use their tool of choice to interact with it.
[10:00:50] <toasterson> I am curious about feedback about it. Is it still interesting to you? What would you like to see implemented?
[10:03:08] <Agnar> hi toasterson, I'm in Bern till friday morning...shall we meet and have a tea? ;)
[10:04:12] <toasterson> Agnar (IRC): Yeah gladly. Want to visit the Keyburg Hackerspace aswell?
[10:04:46] <Agnar> sure, are there any parking lots nearby? I'm currently close to main station and have my car parked in the City West
[10:06:22] <toasterson> Yes it has parking spaces. I am in Breitenrain area. And there is usually Parking spaces infront of the Buildings and on the side of streets.
[10:06:44] <Agnar> cool, let me look i tup
[10:07:13] <toasterson> https://www.chaostreffbern.ch/en/hackerspace.html
[10:08:43] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[10:10:10] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[10:11:49] <ptribble> tsoome: I hit 12322 last night, with the same change to get the build going again
[10:14:00] <tsoome> ok:)
[10:14:28] *** andy_js <andy_js!~andy@> has joined #illumos
[10:16:02] <tsoome> ptribble https://github.com/illumos/illumos-gate/compare/master...tsoome:build has latest push now, will see how the nightly will behave now
[10:19:16] <ptribble> I'm trying a SPARC build with my patches at the moment
[10:20:05] <ptribble> The primary message from that build so far is that I need to build a new release as the build host is starting to get too old
[10:23:49] <tsoome> yes, that is true, I had to hack a bit my m20
[10:26:15] <tsoome> my next issue is that I need to get ability to perform onu
[10:28:02] <ptribble> I ought to write something equivalent fro my packages
[10:28:02] <tsoome> I am starting to get cleanish builds now and need to have ability to boot with new bits.
[10:29:25] <tsoome> still my python 3.8 build is also not quite good, nightly is complaining: Consider setting $PYTHONHOME to <prefix>[:<exec_prefix>]
[10:31:33] *** gmodena <gmodena!~gmodena@> has joined #illumos
[10:38:00] <tsoome> only 9 patches to get gcc 8 done (not counting values*.o issues)
[10:43:06] <Agnar> toasterson: but not today, I have a call tonight I can't miss. is tomorrow fine? what time?
[10:44:00] <toasterson> Yes definetly there will also be others in the space. And we have Teas MATE and other Drinks
[10:44:11] <Agnar> toasterson: cool
[10:45:13] <toasterson> Agnar (IRC): If you have some Redhat or other Stickers bring them. We have a Sticker exchange.
[10:48:46] <Agnar> toasterson: I doubt that I have any ;)
[10:49:20] <toasterson> Agnar (IRC): no worries. Just on the off chance you had.
[10:51:10] *** man_u <man_u!~manu@manu2.gandi.net> has joined #illumos
[10:57:41] <toasterson> Agnar (IRC): at 19:00?
[11:07:00] *** _alhazred <_alhazred!~Alex@> has joined #illumos
[11:33:16] <andyf> tsoome fantastic! and I think my proposed gcc 9.2 branch is good for the values issues once 12306 lands
[11:33:30] <andyf> (we can backport to gcc 8 if it helps you)
[11:39:15] <tsoome> I do not think we should worry about gcc 8 in that sense - we should just skip it. I used it because it was latest one at that time.
[11:39:40] *** _alhazred <_alhazred!~Alex@> has quit IRC ()
[11:51:26] <ptribble> Having cleaned up my SPARC build so it doesn't immediately blow up, now I get into all the ld guidance errors
[11:55:48] <Agnar> toasterson: klingt gut
[12:02:57] <jlevon> ptribble: sorry ;)
[12:04:54] <ptribble> Just looking at the library names it's complaining about, it seems to be mostly around picl, so it might be fairly localized
[12:05:37] <andyf> the ld guidance stuff is finding real problems though - some interesting ones
[12:10:21] <v_a_b> toasterson Come on over to #illumos-de :-)
[12:48:47] *** hemi770 <hemi770!~hemi666@unaffiliated/hemi770> has joined #illumos
[12:53:21] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[12:57:04] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 265 seconds)
[13:07:02] *** sjorge <sjorge!~sjorge@unaffiliated/sjorge> has quit IRC (Quit: 410 Gone)
[13:11:05] *** joltman <joltman!znc@gateway/vpn/privateinternetaccess/joltman> has quit IRC (Ping timeout: 272 seconds)
[13:11:19] *** joltman <joltman!znc@gateway/vpn/privateinternetaccess/joltman> has joined #illumos
[13:16:35] *** sjorge <sjorge!~sjorge@unaffiliated/sjorge> has joined #illumos
[13:30:30] <toasterson> v_a_b (IRC): oh good to know will do.
[13:52:42] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[13:56:11] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 258 seconds)
[14:00:11] *** tru_tru <tru_tru!~tru@> has joined #illumos
[14:23:05] *** CustosLimen <CustosLimen!sid53872@gateway/web/irccloud.com/x-tuvbzcmgrlykxmbb> has joined #illumos
[14:26:14] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[14:29:52] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 265 seconds)
[14:34:55] *** patdk-lap <patdk-lap!~patrickdk@> has joined #illumos
[14:38:46] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[14:39:27] <leoric> Need one more eyes on http://buildzone.oi-build.r61.net/webrev-12316/
[14:42:35] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 268 seconds)
[14:42:38] <gitomat> [illumos-gate] 9794 remove bogus casts from PCI_CAP_PUTx and PCI_XCAP_PUTx definitions -- Hans Rosenfeld <hans.rosenfeld at joyent dot com>
[14:52:52] <tsoome> ptribble if you could try out https://code.illumos.org/c/illumos-gate/+/383 - I can create format-patch if needed.
[14:53:19] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has joined #illumos
[14:58:01] <ptribble> tsoome: thanks, I've just grabbed the patch, hoping to find some more time tomorrow to test it
[14:59:25] <tsoome> I just have to wait 2 hours for my T4-1 to complete the build:D
[15:01:41] <ptribble> Ha. Quicker than my T5140.
[15:02:20] <ptribble> Although my builds were much quicker today, basically because they all failed.
[15:02:29] <Agnar> heck, a good reason to buy a T4-2...it's just that is has too few disk slots
[15:09:53] <v_a_b> Agnar Get a cheap NAS and export iSCSI LUNs. I use an HP Microserver N40L with SmartOS for that.
[15:11:37] <Agnar> v_a_b: the keyword here is "housing costs"
[15:44:45] <andyf> I have a weird possible linker problem..
[15:45:31] <andyf> For crt1.o, recent binutils creates relocation data using R_AMD64_PC32 instead of R_AMD64_PLT32
[15:45:45] <andyf> https://paste.ec/paste/1fooe7qk#M9my3Yhh4r6OmYboidh93hzVFhqfe92IS+4BXWYJ0Nz
[15:46:08] <andyf> and then when building shared objects with -G, ld.so.1 says
[15:46:09] <andyf> 18205: 1: ld.so.1: conftest: fatal: relocation error: R_AMD64_PC32: file ./conftest.so: symbol _start_crt: value 0x3801109fe18 does not fit
[15:46:42] <andyf> (well, when subsequently opening such shared objects)
[15:49:07] *** tsoome <tsoome!~tsoome@148-52-235-80.sta.estpak.ee> has quit IRC (Ping timeout: 260 seconds)
[15:49:34] *** nde <nde!uid414739@gateway/web/irccloud.com/x-tkixvnweygpowdif> has joined #illumos
[15:49:51] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has joined #illumos
[15:51:53] <Agnar> andyf: I'd love to help, but I have no idea how linkers work in detail :)
[15:52:09] <Agnar> I could need some training on that ;)
[15:52:15] <andyf> Me neither :) but I did find a commit in binutils/gas where they disable PLT32 relocations
[15:52:21] <andyf> (for Solaris)
[15:55:00] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has quit IRC (Ping timeout: 248 seconds)
[15:55:37] <jlevon> andyf: well that's going to be fun to figure out
[15:56:03] <andyf> jlevon: it's been fun just getting this far... (assuming the same interpretation of fun..)
[15:56:11] <andyf> at least I was able to eliminate gcc7 early :)
[16:00:03] <jlevon> I presume that commit message gives you no clue as to why they disabled PLT32
[16:00:42] <andyf> yes, but it says that PLT32 is not supported on solaris by krtld nor ld.so
[16:00:43] <andyf> https://github.com/bminor/binutils-gdb/commit/a5def729be2596496aec225e843903b25c672e01
[16:04:36] *** tru_tru <tru_tru!~tru@> has quit IRC (Ping timeout: 265 seconds)
[16:06:44] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has joined #illumos
[16:18:21] <jlevon> andyf: not sure that's the issue
[16:18:52] <jlevon> the gas patch should fall back to what was there already in the changed case
[16:25:24] *** tsoome <tsoome!~tsoome@148-52-235-80.sta.estpak.ee> has joined #illumos
[16:27:52] <andyf> reverting that patch ends up with PLT32 back in the relocation section
[16:28:22] <andyf> but I don't know if that's right, it's just what it was before
[16:28:25] <jlevon> andyf: which one? rainer's or GJ's?
[16:28:27] <jlevon> HJ's
[16:28:45] <andyf> Rainer's
[16:30:00] <andyf> What does this show on smartos?
[16:30:01] <andyf> elfdump -N .rel.text /usr/lib/crt1.o| grep '_crt$'
[16:30:13] <jlevon> ok so I guess it's a symptom of something else.
[16:30:24] <andyf> or the /amd64/ version actually
[16:31:13] <jlevon> you mean .rela.text presumably
[16:31:21] <andyf> elfdump -N .rela.text /usr/lib/amd64/crt1.o | grep '_crt$'
[16:31:21] <andyf> sorry
[16:31:27] <jlevon> R_AMD64_PC32 0x14 0xfffffffffffffffc .rela.text _start_crt
[16:32:09] <jlevon> PLT32 on my oi box
[16:32:20] <andyf> it seems to have changed with binutils 2.33
[16:32:40] <andyf> and the symptom that I was tracking down is that 'zsh' configure now says we don't support dynamic modules
[16:33:30] <jlevon> we have binutils-2.25
[16:36:40] <jlevon> I don't know offhand how we end up with PC32 in smartos
[16:38:25] <andyf> no, or if it even matters - the real problem is probably elsewhere
[16:40:05] <jlevon> put the bad lib up somewhere?
[16:40:50] <jlevon> the error is really making it sound like _start_crt sym value is nuts
[16:41:14] <andyf> the crt1.o or the library that ld.so.1 can't process?
[16:41:39] <jlevon> both I guess
[16:42:37] <jbk> the commens in doreloc.c and machrel.amd.c would seem to suggest we do support PLT32 (and have for quite some time)..
[16:42:54] <jlevon> yeah I can't make much sense of rainer's comment
[16:43:12] <andyf> https://downloads.omnios.org/misc/ld/
[16:43:16] <andyf> no, me neither
[16:43:34] <andyf> lib.G.so is the one built on an older system
[16:43:52] <andyf> lib.G.r30.so even <- this is the old, working one
[16:44:00] <andyf> lib.G.so has the relocation problem
[16:44:04] <jbk> we don't support R_AMD64_GOT32.. maybe he mixed them up??
[16:44:21] <jbk> (at least in ktrld.)
[16:52:57] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has quit IRC (Quit: Leaving)
[16:53:07] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has joined #illumos
[16:53:12] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has quit IRC (Read error: Connection reset by peer)
[16:57:08] <jlevon> hmm wait
[16:57:20] <jlevon> why is crt1 in the lib
[17:00:59] <andyf> it looks like the linker is stripping the PLT32 relocation from .SUNW_reloc in the final .so object
[17:01:06] <andyf> but not the PC32 one
[17:01:21] <jlevon> what's your full link line?
[17:01:52] <andyf> https://paste.ec/paste/x50qJoYS#lNdzqc-+aoXETA2N37ts+ZR3MnCsnP/UxsC/jN7Quou
[17:02:13] <jlevon> why are you adding crt1 there
[17:02:25] <andyf> I have no idea, gcc is doing it
[17:03:12] <jlevon> it created a /tmp/crt1.o ?? crikey
[17:03:17] <andyf> ah no sorry
[17:03:26] <andyf> I put the version in /tmp
[17:03:52] <jlevon> ok can we go back a step to the point where gcc decided to include crt1
[17:03:55] <andyf> gcc uses /usr/lib/amd64/crt1.o
[17:07:30] <andyf> It is because a lot of autoconf stuff builds shared libraries by calling gcc with just -G
[17:07:36] <andyf> (for platform = solaris*)
[17:08:02] <andyf> If they added `-shared`, then crt1.o would not be included
[17:08:46] <jperkin> I confess to patching a bunch of stuff in pkgsrc to s/-G/-shared/ without really understanding exactly what the root cause is of why it fixes it, but it does
[17:08:50] <andyf> This works (for some value of...) if crt1.o has a PLT32 relocation in it
[17:09:02] <andyf> but not if PC32 is in there
[17:10:55] *** robertdfrench <robertdfrench!~robertdfr@h69-11-139-135.cncrtn.dedicated.static.tds.net> has joined #illumos
[17:12:17] <andyf> jperkin: I see the exact patch for zsh :) https://github.com/NetBSD/pkgsrc/blob/trunk/shells/zsh/patches/patch-configure.ac#L98
[17:12:48] <andyf> which is what started me looking at this, since we had a regression in the zsh build following a binutils update
[17:13:53] <jperkin> yeh, if you're able to get to the bottom of why -G is broken that'd be awesome
[17:14:01] <jperkin> sorry, just one of those things I didn't have time for :/
[17:14:40] <jlevon> building w/o -shared was always mildly broken, I think.
[17:17:12] <andyf> We could probably patch gcc...
[17:19:50] <jlevon> it's unclear what an undocumented -G even really means
[17:20:37] <andyf> It's in the man page - it means..
[17:20:38] <andyf> -G Create a shared object. It is recommended that -symbolic or
[17:20:38] <andyf> -shared be used instead.
[17:21:16] <jlevon> which man page
[17:21:21] <andyf> gcc
[17:21:45] <andyf> You have to go down to "Options for System V"
[17:21:48] <jlevon> hmm yeah nm
[17:21:51] <jlevon> broken search string
[17:22:13] <jlevon> so the proximal thing here, I think, is that we have always had a broken -G option
[17:22:32] <jlevon> it can't make sense to have a crt1 in a shared lib afaik
[17:22:43] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[17:24:02] <andyf> ok, we apparently just got away with it before, depending on how gas wrote out the relocation information
[17:24:14] <andyf> but who knows how correct it was..
[17:24:20] <jlevon> yeah I think so
[17:24:24] <jlevon> but not sure
[17:24:39] <andyf> and to be clear, it's the gcc -G option.. in ld, -G is exactly equivalent to -shared
[17:25:20] <jlevon> right
[17:25:38] <jlevon> hopefully specs will allow us to fix up -G to mean -shared
[17:25:52] <andyf> Thanks, that's been really helpful - now to find where gcc implements -G
[17:25:57] *** neuroserve <neuroserve!~toens@> has quit IRC (Ping timeout: 260 seconds)
[17:26:15] <jlevon> andyf: I'm hoping specs is sufficient!
[17:26:25] <jlevon> also seems like an upstreamable fix
[17:26:30] <andyf> I would have thought that -G would set shared to true for specs
[17:26:35] <andyf> but apparently not
[17:26:42] <jlevon> 550743: symbol=_start_crt; lookup in file=a [ ELF ]
[17:26:42] <jlevon> 550743:
[17:26:42] <jlevon> 550743: file=./lib.G.so; modifying memory protections (+ PROT_WRITE)
[17:26:42] <jlevon> ld.so.1: a: fatal: relocation error: R_AMD64_PC32: file ./lib.G.so: symbol _start_crt: value 0x4002371fdd8 does not fit
[17:26:55] <jlevon> presumably they literally just pass -G through to the linker
[17:27:34] <jlevon> that's the LD_DEBUG output. so, at a guess, now we're finding _start_crt symbol up in the executable and that happens to be too far away to fit in the PC32
[17:28:12] <jlevon> 550743: file=./lib.G.so [ ELF ]; generating link map
[17:28:12] <jlevon> 550743: addr: 0xfffffbffdcce0000 size: 0x1164d
[17:28:18] <jlevon> 550743: file=/home/gk/a [ ELF ]; generating link map
[17:28:19] <jlevon> 550743: addr: 0x400000 size: 0x11830
[17:28:28] <jlevon> seems to make sense
[17:28:59] *** robertdfrench <robertdfrench!~robertdfr@h69-11-139-135.cncrtn.dedicated.static.tds.net> has quit IRC (Quit: WeeChat 2.6)
[17:31:56] <jlevon> not quite sure how we get to 0x4002371fdd8 though
[17:32:15] <andyf> PC32 is relative to something..
[17:32:28] <jlevon> every time I've followed a relocation calculation though, it's ended up making sense eventually haha
[17:33:05] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[17:33:31] <jlevon> well in this case, it's pc relative no?
[17:34:23] <jlevon> so it would be loaded address of lib.G.so`_start+0x13f4 as the pc
[17:35:36] <jlevon> yep
[17:35:42] <jlevon> >>> hex(0xfffffbffdcce0000 + 0x13f4)
[17:35:42] <jlevon> '0xfffffbffdcce13f4L'
[17:35:42] <jlevon> >>> hex(0xfffffbffdcce13f4 + 0x4002371fdd8)
[17:35:42] <jlevon> '0x100000000004011ccL'
[17:35:54] <jlevon> $ nm a | grep start_crt
[17:35:54] <jlevon> 00000000004011d0 T _start_crt
[17:35:58] <andyf> a bit too big for 32-bits?
[17:36:18] <jlevon> it's PC32 so at some point the masking will come in. forget exact details
[17:36:59] <jlevon> modulo another 4 bytes that marries up
[17:37:58] <jlevon> what I still don't follow is why this broke now. from that patch, rainer's change should have just left us back at the original version
[17:38:36] <andyf> no, me neither. 2.32 was outputting PLT32, 2.33 does PC32
[17:38:41] <andyf> and your 2.53 does PC32 too
[17:38:48] <andyf> or 2.52, whatever it was
[17:39:13] *** alanc <alanc!~alanc@inet-hqmc02-o.oracle.com> has quit IRC (Remote host closed the connection)
[17:39:41] *** alanc <alanc!~alanc@inet-hqmc02-o.oracle.com> has joined #illumos
[17:40:12] <jlevon> maybe it's something like, some previous change in between caused gas to emit PC32
[17:40:30] <jlevon> *then* hj's patch came in to translate those to PLT32 for whatever reason
[17:40:42] <jlevon> *then* rainer's patch disabled that for us again, and we're back to PC32
[17:40:44] <andyf> yes, binutils 2.32 could be in that gap
[17:41:01] <andyf> it's interesting that the linker removes the PLT32 things from SUNW_reloc
[17:41:03] <jlevon> maybe smartos binutils is so old it was PC32 again
[17:41:09] <andyf> but definitely seems to be a gcc problem
[17:41:50] <andyf> and gcc is not the most greppable source tree I've looked at..
[17:43:39] <jlevon> 260 "%{G:-G} \
[17:43:45] <jlevon> in gcc/config/sol2.h
[17:43:54] <jlevon> I wonder if that can be %{G:-G -shared}
[17:44:00] <jlevon> dunno if it expands recursively
[17:44:35] <andyf> or we could tweak STARTFILE_SPEC
[17:44:35] <andyf> #define STARTFILE_SPEC "%{!shared:%{!symbolic: \
[17:44:35] <andyf> crt1.o%s \
[17:44:42] <andyf> Adding %{!G
[17:44:50] <jlevon> yeah
[17:44:57] <jlevon> that seems like what we need
[17:45:28] *** wonko <wonko!~quassel@> has quit IRC (Remote host closed the connection)
[17:45:58] *** wonko <wonko!~quassel@> has joined #illumos
[17:46:27] <andyf> I'll give it a go, thanks for the help!
[17:47:47] <andyf> If it seems to work, I can add it to my draft gcc9.2 PR
[17:48:08] <jlevon> awesome
[17:48:49] <andyf> I'm just struggling with processing test results on that branch, otherwise it would be further forward - but I'll get back to it, hopefully in time for tsoome to jump onto gcc9
[17:50:43] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[17:53:56] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 248 seconds)
[18:03:01] <jlevon> https://gist.github.com/jlevon/0dcda36103380125f59aa2d6c80d3027
[18:03:26] <jlevon> who knows why the difference
[18:14:13] *** BOKALDO_ <BOKALDO_!~BOKALDO@> has joined #illumos
[18:14:36] *** BOKALDO <BOKALDO!~BOKALDO@> has quit IRC (Read error: Connection reset by peer)
[18:18:16] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has joined #illumos
[18:21:23] *** neirac_ <neirac_!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Ping timeout: 260 seconds)
[18:24:45] <jlevon> andyf: very recent 292676c15a615b5a95bede9ee91004d3f7ee7dfd seems relevant too
[18:27:00] <rmustacc> What's going on? Do we need to add support for some newer relocation types?
[18:27:30] <jlevon> summary so far
[18:27:52] <jlevon> newer binutils gas generates PC32 for the crt1 call to _start_crt (local sym)
[18:28:10] <jlevon> it used to be PLT32 ( though given that it's not call _start_crt@PLT, I'm not sure why)
[18:28:38] *** man_u <man_u!~manu@manu2.gandi.net> has quit IRC (Ping timeout: 245 seconds)
[18:28:39] <jlevon> then, as it happens, gcc -G doesn't translate to the -shared option, and in our gcc specs, we end up including crt1.o in that case.
[18:29:20] <jlevon> then when we build an executable against such a shared library, we try to resolve libfoo.so`_start's relocation call of _start_crt to the one far away in a.out`crt1.o
[18:29:27] <jlevon> and that can't fit into the PC32 relocation
[18:29:34] <jlevon> sorry, run an executable not build.
[18:29:54] <jlevon> proximal fix is to make -G the exact same as -shared (since crt1.o shouldn't be in a shared lib)
[18:30:18] <tsoome> does this one provide hints http://gcc.1065356.n8.nabble.com/relocation-error-R-AMD64-PC32-td649177.html ?
[18:30:55] <rmustacc> Did -G ever mean -shared in gcc?
[18:31:12] <tsoome> "GCC won't pass options through to the linker, and -G means something
[18:31:12] <tsoome> else to GCC, it probably isn't passed to the linker at all. "
[18:31:22] *** wiedi <wiedi!~wiedi@> has quit IRC (Ping timeout: 268 seconds)
[18:31:31] <rmustacc> Right, I think us passing -G to gcc unless as -Wl,-G is probably erroneous?
[18:31:40] <rmustacc> So, it should probably always be -shared?
[18:31:46] <rmustacc> Though all the other bits are also troublesome.
[18:31:48] <tsoome> seems so, yes
[18:32:09] <jlevon> rmustacc: it's explicitly documented, though, and also legion
[18:32:28] <jlevon> this is a legacy thing and would involve fixing a lot of software.
[18:32:30] <rmustacc> jlevon: It is? All the docs I could find right now don't say that at least.
[18:32:34] <rmustacc> OK.
[18:33:04] <jlevon> it's way down in the man page
[18:33:10] <jlevon> -G Create a shared object. It is recommended that -symbolic or -shared be used instead.
[18:33:18] *** edef <edef!edef@NixOS/user/edef> has quit IRC (Ping timeout: 260 seconds)
[18:33:54] <rmustacc> I see, I found that eventually in the System V option section.
[18:34:03] *** edef <edef!edef@NixOS/user/edef> has joined #illumos
[18:34:22] <jlevon> it is nuts this dates back to 2011 as tsoome found
[18:34:51] <rmustacc> Wow.
[18:36:58] <rmustacc> So I guess the compiler -G thing is primary and secondary is probably thinking about the reloc stuff I guess and if that's something we need to think about.
[18:38:12] <jlevon> yes
[18:38:34] <jlevon> the history is a bit confusing to me as to PC32 versus PLT32
[18:38:48] <jlevon> and I'm inclined not to worry about it. this should never have been a PLT rel in the first place
[18:39:07] <jlevon> regardless of whether and when plt32 got generated
[18:44:57] <rmustacc> OK.
[18:45:13] <rmustacc> So it sounds like there isn't an rtld bug we need to address?
[18:45:45] <jlevon> I don't think so
[18:46:01] <rmustacc> OK. Gotcha.
[18:46:20] <jlevon> the one niggle, maybe, is whether the patch that got disabled in gas might provide some advantage and we should in fact re-enable it. not clear on that at all.
[18:47:09] <jlevon> hopefully andyf will have success with the gcc specs fix and we can pull that in everywhere
[18:47:18] <jlevon> and maybe try upstreaming it
[18:48:06] <rmustacc> Gotcha. OK. Thank you for the summary.
[18:56:11] *** _alhazred <_alhazred!~Alex@> has joined #illumos
[19:48:49] *** trisk <trisk!~trisk@2601:196:4700:3a:8:20ff:fe3e:890e> has quit IRC (Ping timeout: 272 seconds)
[19:50:08] *** trisk <trisk!~trisk@2601:196:4700:3a:8:20ff:fe3e:890e> has joined #illumos
[19:56:51] *** neuroserve <neuroserve!~toens@ip-178-202-216-248.hsi09.unitymediagroup.de> has joined #illumos
[19:58:27] <neuroserve> trying to compile minio in a smartos-zone: package github.com/minio/minio: build constraints exclude all Go files in /root/go/src/github.com/minio/minio ?!
[20:02:07] *** trisk <trisk!~trisk@2601:196:4700:3a:8:20ff:fe3e:890e> has quit IRC (Ping timeout: 272 seconds)
[20:11:37] *** trisk <trisk!~trisk@c-24-91-168-164.hsd1.ma.comcast.net> has joined #illumos
[20:16:55] *** wiedi <wiedi!~wiedi@ip5b4096a6.dynamic.kabel-deutschland.de> has joined #illumos
[20:32:20] <bahamat> Is that an error message you're getting?
[20:32:47] <bahamat> neuroserve: ^^
[20:32:59] <papertigers> neuroserve: this may be helpful to you
[20:33:00] <papertigers> https://github.com/omniosorg/omnios-extra/tree/master/build/minio
[20:38:30] <LeftWing> neuroserve: What version of Go are you using?
[20:38:56] <LeftWing> ("go version" I think)
[20:41:39] <neuroserve> LeftWing: that error came with 1.12 - when I use go1.13.5 the error goes away - but it does not compile...
[20:48:27] <LeftWing> That error is expected, I think, with anything less than 1.13
[20:48:42] <LeftWing> https://github.com/minio/minio/blob/master/main.go#L1
[20:48:49] <LeftWing> The build tag in this file requires 1.13
[20:49:00] <LeftWing> Presumably you'll get some other error(s) when building with >=1.13
[20:49:09] <LeftWing> Which will be more actionable with patches/fixes
[20:50:36] *** cneir__ <cneir__!~cneir@pc-184-104-160-190.cm.vtr.net> has quit IRC (Quit: Leaving)
[20:53:25] *** trisk <trisk!~trisk@c-24-91-168-164.hsd1.ma.comcast.net> has quit IRC (Ping timeout: 240 seconds)
[20:54:47] *** trisk <trisk!~trisk@2601:196:4700:3a:8:20ff:fe3e:890e> has joined #illumos
[20:55:44] *** bahamat_ <bahamat_!~bahamat@> has joined #illumos
[20:59:43] *** bahamat_ <bahamat_!~bahamat@> has quit IRC (Client Quit)
[20:59:56] *** bahamat_ <bahamat_!~bahamat@> has joined #illumos
[21:01:20] <Smithx10> LeftWing: dope vendor id
[21:01:35] <LeftWing> We were certainly very pleased with ourselves haha
[21:01:42] <Smithx10> Anymore coming down the pipe?
[21:03:54] <Agnar> LeftWing: indeed, very cool :)
[21:04:38] <Smithx10> Agnar: I hope they put all the requests in... I can imagine after that tweet someone might attempt to snipe one
[21:05:10] <Agnar> Smithx10: indeed, but those guys are awesome, they'll get it done :)
[21:06:50] <Smithx10> Wonder if we will ever see the "OxProc"
[21:07:12] <Smithx10> or..... "Ox86"
[21:07:17] <Smithx10> ./walks away slowly
[21:07:50] <Agnar> haha
[21:25:13] *** BOKALDO_ <BOKALDO_!~BOKALDO@> has quit IRC (Quit: Leaving)
[21:25:43] *** bahamat <bahamat!~bahamat@> has quit IRC (Quit: ZNC - http://znc.in)
[21:25:52] *** bahamat_ <bahamat_!~bahamat@> has quit IRC (Quit: ZNC - https://znc.in)
[21:26:09] *** bahamat <bahamat!~bahamat@> has joined #illumos
[21:29:08] *** kayront <kayront!~kayront@unaffiliated/kayront> has joined #illumos
[21:41:10] <sjorge> Doesn't intel have few close to 0x86
[21:50:19] *** jimklimov <jimklimov!~jimklimov@ip-86-49-254-26.net.upcbroadband.cz> has quit IRC (Quit: Leaving.)
[22:24:36] <jbk> i thought they had 0x8086
[22:29:07] <alanc> yeah, that's Intel's PCI vendor id
[22:32:52] <jbk> and in a bit of coincidence, both me and someone else i know ended up getting that for the last 4 digits of our cell#s :)
[22:39:35] <alanc> I see it's made pci.ids now: http://pci-ids.ucw.cz/read/PC/01de
[22:44:22] <sjorge> jbk ah yeah it migt inded be 8086
[22:45:02] <sjorge> I wonder what there mac range will be :p it better be good
[22:49:47] <sjorge> for https://www.illumos.org/issues/12317 I assume it's not going to be as easy as moving "case I40E_PHY_TYPE_SGMII:" up to line 1220 in http://src.illumos.org/source/xref/illumos-gate/usr/src/uts/common/io/i40e/core/i40e_common.c#1243
[22:52:28] <sjorge> I mean... I will know tomorrow xD as I am doing a PI build rn
[22:52:29] *** kahiru <kahiru!~quassel@ip-89-102-207-18.net.upcbroadband.cz> has quit IRC (Ping timeout: 272 seconds)
[22:53:25] *** kahiru <kahiru!~quassel@ip-89-102-207-18.net.upcbroadband.cz> has joined #illumos
[22:59:52] <jbk> i don't know offhand myself
[22:59:59] <jbk> i'd have to read the programming guide
[23:05:09] <sjorge> This is so out of depth for me
[23:05:20] <sjorge> Also specs suck
[23:05:43] <jlevon> "i40e: the ring cycle"
[23:05:52] <sjorge> X722 oh no we don't do SGMII because no compatiple PHY suppors that, X577 PHY: hold my beer
[23:06:06] <sjorge> So far I only every got a link with only 1 cable plugged in on windows
[23:06:35] <sjorge> And illumos with 2 plugged in and restting the box while the links are up switch side (other startup mode for the device)
[23:07:04] <sjorge> Anyway, throwing a tiny pebble at it should be hilarious and not effective I think
[23:07:23] <sjorge> But I was testing gmake clean world stuff after @jlevon fix for smartos-live so itseemed like a fun thing to try
[23:21:21] *** andy_js <andy_js!~andy@> has quit IRC (Quit: andy_js)
[23:36:58] *** ptribble <ptribble!~ptribble@cpc92716-cmbg20-2-0-cust138.5-4.cable.virginm.net> has quit IRC (Quit: Leaving)

   February 19, 2020  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | >