IRC logs for #openrisc Saturday, 2014-05-03

--- Log opened Sat May 03 00:00:31 2014
stekernpoke53282: yes, running scummvm X in a vncserver on a FPGA05:33
stekernpoke53282: hmm, qemu crashes when I try to build it05:46
stekernor well, gcc crashes... and it does it on the board too05:59
stekernoh, this might be the mpfr bug, I probably have an old version of that06:02
stekernnow it compiles06:02
stekernpoke53282: mor1kx@75MHz:
stekernas a comparison, this is what we get in qemu-user:
blueCmdstekern: apparently apache needs support for -pie for gcc07:22
blueCmdor I'm just stupid, stand bye07:23
blueCmdno, debian uses --enable-pie with apache2. I will need to remove that for or1k07:24
blueCmd(or implement it, not sure what the differences are yet between that and pic though)07:24
blueCmd"These options are similar to -fpic and -fPIC, but generated position independent code can be only linked into executables. Usually these options are used when -pie GCC option is used during linking.07:24
stekernumm, ok... what's the problem?07:33
stekernwe support -PIE afaik07:33
stekernat least I have used it in the past, so we *have* supported it07:34
stekern-fPIE I mean07:34
blueCmdi have done 0 debugging though, im traveling for 2 weeks07:36
stekernok, so --enable-pie is a ./configure flag07:37
blueCmdmaybe it just tests -pie and not fPIE07:38
stekernmmm, or there's something more subtle going on07:39
blueCmdhard to say without the config.log07:39
stekernindeed ;)07:41
stekernI can try-build it here and see what it tries to do07:46
stekernright after this wget build is done ;)07:46
blueCmdcool, thanks07:46
blueCmdwget is built07:46
stekernis it?07:47
blueCmdyou might want to add sid main contrib to your sources07:47
stekernI couldn't apt-get install it this morning07:47
stekernah, ok07:47
blueCmdthat's where wannabuild puts all it's stuff07:47
blueCmdit's the "pure hack-free" repository07:48
blueCmdsorry if I didn't tell you about that earlier07:48
blueCmdanyway, apt fails 1 test with file related stuff,
blueCmdthat might be interesting to look into why it fails07:48
blueCmd shows an example sources.list07:52
stekernhow do I use that pubkey?07:55
stekernnm, google told me07:59
stekern(apt-key add
-!- Netsplit *.net <-> *.split quits: _franck_, zama, rah, ysionneau, blueCmd, poke53282, veprbl, jeremy_bennett, Limb08:04
-!- Netsplit over, joins: jeremy_bennett, _franck_, blueCmd08:07
stekernblueCmd: I get this when I try to build apache: /usr/share/apr-1.0/build/libtool: line 1088: or1k-linux-gnu-ar: command not found08:21
stekernhow can I not have had binutils until now?08:23
blueCmdstekern: you have binutils, but the old version wasnt packaged in the debian way and was just "ar"09:32
blueCmdbeware that the latest binutils does _not_ have the TLS relocation bug fixed, but it should work for most things09:32
stekernah, ok..09:37
stekernwhy doesn't it? didn't you push that upstream?09:38
blueCmdI did, but it was packaged before I did that09:38
stekerncrystal clear09:38
stekernbut hey...09:38
blueCmdwho would have thought porting Debian would involve a lot of moving parts?09:39
stekernapache built without problems on my machine09:42
stekernno hacks, no voodoo, no nothing09:43
blueCmdstekern: interesting09:44
blueCmdi'll add it for now09:44
blueCmdmind taring? :)09:44
stekernblueCmd: oh, there was a couple of libapache too:
blueCmdstekern: you might find this informative:
blueCmdit's how many packages depend on that package09:59
blueCmd(november 2013, but should be relevant still)10:00
stekernwhy are there so many that have 19554?10:01
blueCmdbecause cycles10:02
blueCmdit's called the "debian blob"10:02
stekernah, I understand10:02
stekernntp is now confirmed to work10:18
blueCmdstekern: did you write the real atomics btw? I could use the .c / .S file as a reference how to implement CMPXCHG and the others10:20
stekernI used the musl implementation:
stekernsome of those could be optimised instead of just using the a_cas implementation10:24
blueCmdyou should probably consider using the gcc internals there later10:24
stekernI don't think that's a good idea, that'd make musl depend on gcc10:26
stekernthis is what I linked into scummvm:
blueCmdright, yes that's true10:27
blueCmdmaybe it's bad to depend on it for glibc as well then10:29
blueCmdit will make an llvm port harder10:29
stekernheh, I think the glibc people consider a gcc dependency a *good* thing though10:30
blueCmdyes, I just googled "llvm glibc" and it appears that it's "one of those" libraries that will probably only work with gcc10:31
stekernbut IMO, lock-ins are always bad, regardless of the motives10:31
blueCmddamn the musl port is small and neat10:32
stekernapache2 doesn't want to start:
blueCmdstekern: ah cool. please file a bug in or1k-debian about "apache2-mpm-event / apr not supporting Thread Safe Pollset"11:15
blueCmdtry using apache2-mpm-prefork instead11:16
stekernsome old bug-reports speak about apache2-mpm-worker11:17
stekernhow do I use any of those?11:17
blueCmdI'm not sure, traditionally you just installed it - but that seems to be depreated11:18
blueCmdbut try to install apache2-mpm-worker and it should switch11:19
stekernheh.. I wondered why it took so long.. but apt-get install apache2-mpm-worker ended in a kernelpanic11:46
blueCmd:D interesting!11:47
blueCmdmaybe in my kernelspace atomics that doesn't do any sanity  checking11:48
stekernwell, from a debugging perspective it doesn't make much of a difference11:54
stekernblueCmd: re postgresql, I get this:12:29
stekern[warn] No PostgreSQL clusters exist; see "man pg_createcluster" ... (warning).12:29
stekernapart from that, no explosions12:30
stekernmpm-worker doesn't work much better than mpm-event12:37
stekern[Thu Jan 01 00:11:50.092438 1970] [mpm_worker:emerg] [pid 9197:tid 1038480496] (38)Function not implemented: Couldn't create pollset in thread; check system or user limits12:37
GreenHatI am a student new to this community but they want us to run an application on linux and simulate the hole thing using or1ksim, the application is around 20 MB installed, we put it in the support directory of linux under linux/arch/openrisc/support/initframs/ of course merge the direcotries of our application with the existing one. the problem is: or1ksim cannot load the system and run out of memory, so we want to ask which secti21:20
GreenHatanything can help, any suggestions, I came here because there is no or1ksim channel, so I think I can find some developers of this simulator here?21:35
--- Log closed Sun May 04 00:00:33 2014

Generated by 2.15.2 by Marius Gedminas - find it at!