IRC channel logs
2023-01-21.log
back to list of logs
<nektro>just had to do the copy myself, was assuming it'd be automatic coming from other build systems <stikonas>argh, I wanted to do one last force push <fossy>yeah, i disabled force push to master <fossy>so even if you tried by accident it wouldn't go through :P <stikonas>yeah, we definitely want branch protection on master <stikonas>i.e. force pushing to origin instead of some other remote <fossy>yes, relic that never updated <stikonas>fossy: any objections if I artificially add bin and lib symlinks to musl package? <stikonas>that will allow to just unpack musl, unpack python and run "./lib/ld-musl-i386.so.1 ./bin/python3.11" <fossy>stikonas: remove is ok, won't matter at all once new kernel stuff implemented <fossy>yeah i think that musl change is ok, it doesn't hurt <stikonas>doras: is your live-bootstrap pipeline running in a VM? <stikonas>I wonder if that might be the cause for more frequent autoconf/automake intermittent failures <doras>Well... That's a good question. To my understanding it is, yes. <stikonas>perhaps that the key to better reproducing it <stikonas>I was trying to run it in a VM, once I hit it with automakt 1.15 <stikonas>then next run even automake 1.11 caused it <fossy>stikonas: that's very interesting <fossy>stikonas: hmm, i run qubes os on my main dev machine,so everything runs in a "vm", but that is Xen, rather than QEMU.. <fossy>and i've experienced it twice/probably 30 or so full builds in the last month <fossy>(but those don't include automake 99% of the time) <stikonas>I think that VM I was running on and getting failures is also Xen <stikonas>CDPATH="${ZSH_VERSION+.}:" && cd .. && autoconf-2.64 <stikonas>autom4te-2.64: cannot create autom4te.cache: No such file or directory <stikonas>and running "make MAKEINFO=true" again caused the same failure once, but rerunning it once again suceeded <stikonas>I guess autoconf 2.64 has race conditions <stikonas>so might be quite sensitive to specs and settings <stikonas>at least on automake 1.11 it fails building amhello <stikonas>fossy: remove "doc" from SUBDIRS in top-level Makefile.am? <fossy>since that appears to be where the race condition is <fossy>hopefully it doesn't surface anywhere else... <fossy>might have to do a bit more testing <stikonas>well, so far I've only seen in various versions of automake <oriansj>So build single threaded to eliminate race condition? <nektro>ah okay, just wanted to double check <nektro>didnt know if there was an M1 meant to be built by M0 instead <stikonas[m]>M1 is just C version of M0 with a bit of extra features <stikonas>better commnad line parsing, support for features needed by all architectures, etc... <nektro>not sure if that'd be a bug in M0 or hex2 from what ive read i think itd be M0, or im calling it wrong <stikonas>you still need to link resulting file with hex2 <stikonas>nektro: just look at those kaem files in stage0 <stikonas>nektro: we also have stage0 running on UEFI on amd64 <mihi>Is this related to one of the platform-specific M0 versions that confused a label inside a string literal? <mihi>As it is, it does not cause any bugs (of course; you could always hex-encode all strings) - it just looks weird to me (no pun intended) <stikonas>mihi: I guess try removing it and see if anything breaks <mihi>in case the code is needed, it *would* have been buggy, though: It checks escaped whitespace but not escaped colons \x3A :) <stikonas>if you can get tid of that one, it would be even better <stikonas>but I think I saw some strings were encoded by it <mihi>I believe the "weird string" code as such is needed unless you make sure there are no "weird strings" in M2-Planet's source code. <mihi>I was only surprised about that special case with whitespace and colon <stikonas>e.g. blood-elf-0.M1 has ' 0A 3A 45 4C 46 5F 65 6E 64 0A 00'