IRC logs for #openrisc Tuesday, 2015-01-20

--- Log opened Tue Jan 20 00:00:04 2015
-!- E11fangirl is now known as Nala_04:30
-!- Nala_ is now known as E11fangirl04:30
stekernolofk: yeah... that is a bug in cgen, and I've proposed a patch that fixed it05:13
stekern(a patch to cgen that does the same as what now is committed to the generated file as a workaround)05:15
stekernbut the cgen maintainer never seemed to be pleased with it05:16
olofkaha, so the binutils version is changed manually05:26
olofkFound your mail on the cgen ml05:28
olofkhmm... failed to install the or1k-elf tc again08:47
olofkWhat on earth am I doing wrong?08:47
poke53282You decided to compile software, which are a historic mess with tons of obsolete workarounds and bad design decisions?09:00
olofkpoke53282: You might be right. I will file a bug report for that09:01
olofk[PATCH gcc] Fix bad design decisions09:02
olofk[PATCH gcc] Remove obsolete workarounds09:02
ams[BUG ams] Not long enough beard09:03
olofk[BUG olofk] Too much food ends up in beard09:04
amsmm.. food09:04
amsbeard food09:04
olofkmm.. beard beer09:04
olofkWho put the beer in beard09:04
amswho put the beard in the beer sounds worse ..09:05
olofkwallento: Still have problems with upstream newlib. Using your git repo works fine09:05
wallentohey, okay, I will check09:06
wallentoat some point the several for-upstream and master branches got confusing :)09:06
olofkI just did a diff against newlib-2.2.0-1, but I can't find anything that could explain this09:11
olofkOnly a few things regarding (cache?) flushing and some sync_cas stuff that differed09:11
olofkwallento: Pastie here if it tells you anything
wallentowohaa, thats weird09:12
wallentoI will check09:13
wallentoseems like a missing -D__SYSCALL__BLABLA09:13
wallentoI will also do the continuous integration for the upstream then09:14
wallentoand prepare official releases09:14
wallentobut first fix ;)09:14
olofkYeah, it would be great if you could test and use the upstream versions as much as possible09:14
wallentoyeah, got it09:15
wallento  or1k-*-elf | or1knd-*-elf)09:15
wallentonewlib_cflags="${newlib_cflags} -DREENTRANT_SYSCALLS_PROVIDED "09:15
wallentosomehow didn't make it to configure-host09:15
wallentothats weird09:16
wallentoit was in the submitted patch09:16
wallentoi will get back to them09:16
olofkI can change it locally and rebuild to make sure09:18
olofk_franck__: Do you want to apply the fixes to wb_altera_ddr_wrapper?09:22
olofkBonus points if you put it in a separate repo and put a version tag on it09:23
wallentoyeah, I just sent the patch to the list09:25
poke53282lol, got this error during the compilation of openjdk: java.lang.RuntimeException: time is more than 10 years from present: 110453040000009:27
poke53282the problem started on 01-01-2015:
poke53282but it looks like, that I use the newest version.09:28
poke53282What does this tell you about Java?09:28
olofkThat was a fu bug09:30
olofkSeems like they got a fix09:30
olofkI should add that to FuseSoC too. Just add a static file with a date, and check against that when compiling09:31
olofkIf someone uses FuseSoC in 2025 something must be seriously wrong :)09:31
olofkhmm.. now gcc fails to install instead09:33
poke53282Well, we use Fortran77 code here, which would still run on punch cards. Yes, there is something seriously wrong, but no need to disallow compiling.09:33
amsnothing wrong with punch cards .. you need to plan for the zombie invasion09:34
olofkmake[2]: Entering directory '/home/olof/code/or1k/tcnew/bld/gcc/or1k-elf/libstdc++-v3'09:34
olofkmake[2]: *** No rule to make target 'install'.  Stop.09:34
olofkHow is this even possible? Did someone steal the isntall target while I was compiling?09:36
amsbuilding in source tree?09:37
amsdidn't run configure?09:37
amsmake -jN where N is big?09:37
olofkHm.. could be -j509:37
amsdo a make configure-host09:38
amsor some such ..09:38
amsmake configure-target09:38
poke53282ams: the code has some nice statements like "rewind 3", which means that the unit 3 with punch cards (which contains data) should be reset.09:38
olofkI noticed that my script didn't clean out the gcc build directory between the stage 1 and stage 2 builds. That could be a reason09:38
amspoke53282: nod, i've done punch card programming .. even debugged them and taped them ..09:39
poke53282In C code, this is a fseek command.09:39
ams(i like punch card programming ..)09:42
poke53282ams: you must be old.09:42
amsmany people think that09:43
olofkWell ams also likes wind-up grammophones :)09:43
amsolofk: how did you know?09:43
amsseriously ... gf loves 'em, so do i .. i fix them for her when she finds one09:43
_franck__olofk: "separate repo" ?09:48
olofk_franck__: Yes. I've started to realize that it's starting to become pretty messy to have a lot of RTL code in orpsoc-cores09:49
olofkBut it's fine to just apply the patches to orpsoc-cores if you want09:50
_franck__at least it is centralized09:50
_franck__because if github/fjullien desapear, wb_altera_ddr_wrapper would desapear too09:51
olofkDon't scare us like that :)09:51
olofkYes, that's a benefit of having it centralized, but if we are worried about code disappearing we could also clone the repo09:53
_franck__I pushed the fix now. I'll move the repo later09:53
olofkcool. No hurry with moving . Just thought I should mention it09:54
stekernolofk: libstdc++ always fail for me if I try to resume a halted build or try to compile it with anything > -j109:59
olofkstekern: Cool. Is that a bug in our code or in upstream?10:01
stekernno idea10:01
stekernI'm not in a hurry, I can wait for -j1 ;)10:01
olofkI guess it's quicker than continiously rebuilding with -j5 :)10:02
wallentoolofk: its upstream now10:02
wallentoshould build from cvs10:02
wallentoI think the git is delayed..10:02
olofkYep Saw the mail10:02
olofkGood thing that they decided on monthly snapshots. Hopefully we will only have to wait a month for a tar ball release10:03
wallentoyeah, but nevertheless people will depend on our build10:06
wallentoI will create a 2.2.0 + or1k release10:06
olofkstekern: Still fails without -j5.10:46
olofkFuck this shit. I'm disabling c++ now10:47
stekernbut did you start the build over from beginning?10:48
olofkYep. And cleaned out the target directory10:49
olofkAnd no other toolchains in the path10:49
olofkok, so it installs now, but fails to find stdio.h instead11:16
wallentonewlib git is also updated11:49
wallentoeverything upstream now11:49
olofkwallento: Where can I find newlib git?11:49
olofkI tried adding the patch to to newlib-2.2.0-1, but it still fails11:50
olofkcool. I'll try that11:50
olofkwallento: Nope. Still fails :(12:02
wallentommh, whats the error?12:02
olofkwallento: I think this might be the relevant part
wallentookay, seems some autotools are missing12:04
wallentoi will check12:04
wallentoI was thinking upstream executes them12:04
wallentoah, got it12:09
wallentonext round12:13
wallentoits a rather simple line, see my new patch12:14
olofkDo you want me to test it first?12:14
wallentoit will work :)12:14
olofkFamouse last words :)12:14
wallentoI also created the continuous integration for this12:14
olofkWith the upstream git repo?12:16
wallentobinutils and newlib12:17
olofkThe git repo isn't updated yet AFAICS13:04
wallentommh, on my computer or1k-gcc doesn't build anymore13:17
wallentopod2man fails13:17
olofkehhmm.... wallento... ehmm.. it... kind of... still doesn't work13:20
olofkor1k-support.h is missing13:21
wallentoare you sure?13:23
wallentoahhhh, shit13:23
wallentothat was a patch that I never submitted13:23
wallentoI think13:23
wallentolet me check13:23
wallentoit is there but cannot be found, correct?13:23
olofk../../../../../../../newlib-git/newlib/libc/sys/or1k/mlock.c:19:26: fatal error: or1k-support.h: No such file or directory #include <or1k-support.h>13:23
wallentookay, let me fix this, but add some new features also13:25
olofkThere are a few fixes in mlock.c13:25
wallentonamely mmu management13:25
olofkDo you want me to try it before you send it upstream? :)13:25
_franck__wallento: new_features == new_problems13:28
wallentoit is only (i|d)mmu_flush13:28
wallentoin assembly13:28
wallentothat is hopefully no new errors13:29
wallentoproblem is I wanted to change something in or1k-headers for this13:29
wallentoby the way there are GPL headers in or1k-headers, for the nops13:29
wallentowe need a blind monkey to rewrite them13:29
olofkWhat? I thought the idea of or1k-headers was to remove the GPL stuff13:30
wallentoyes, but or1k-nops.h was moved there13:37
wallentoI did NOP_PUTC by heart13:37
wallentothe others I didn't remember and was not supposed to look at the preivous one13:37
wallentobranch for-upstream13:38
olofkwallento: olofk the build bot is on it!13:38
olofkwallento: Fail!13:45
olofkLooks like it can't find sys/or1k13:56
wallentoyou're kidding :)13:58
olofkI wish I did13:58
wallentoI still can't build gcc from the repo13:59
wallentodid something change in ubuntu's pod2man recently?13:59
wallentogfdl.pod around line 53: Expected text after =item, not a number13:59
olofkDon't know. Gentoo user here13:59
olofkThat hasn't been a problem in any of the ~30 gcc compilations today :)14:00
poke53282olofk: Since my mainboard died with my Atom N270, I haven't touched gentoo anymore.14:02
poke53282which gcc version you try to compile?14:02
poke53282wallent: Can you compĆ¼ile with --disable-docs ?14:03
wallentommh, its gone now14:04
wallentoeven better :)14:04
wallentoolofk: do it14:34
olofkDo what?14:34
wallentowas just a git hickup14:37
wallentoupstream was not in, i.e., the auto-generated configure14:37
olofkGot to go now, but I should just pull your for-upstream branch and build, right?14:46
wallentofor me it build now14:50
wallentofor a clean tree14:50
wallentoolofk: i just build successfully from for-upstream, upstream binutils and or1k-gcc15:39
wallentoolofk: works now (I assume I meant objcopy ages ago)15:45
olofkwallento: Good job with the last upstreaming hickups, and thanks for looking at the bug report18:26
--- Log closed Wed Jan 21 00:00:06 2015

Generated by 2.15.2 by Marius Gedminas - find it at!