IRC channel logs

2021-04-26.log

back to list of logs

<xentrac>OriansJ: btw I tried running StoneKnifeForth on Linux 5.8.0-48-generic on amd64 (though of course it's an i386 binary) and it works fine despite having only a single ELF segment, which is mapped RWE
<OriansJ>xentrac: of course, it is related to two overlapping segments with different permissions.
<xentrac>hmm, so not like SKF but like httpdito? but standard toolchains produce that kind of thing all the time
<xentrac>(although I guess the segments don't quite overlap, they do consist of the same page)
<gforce_de1977>stikonas: fossy: i'am getting mad: after adding debug-output to 'sysa/helpers.sh' the are ZERO errors during my massrun. now 640 full build with success and counting....what is interesting: we have some differences in the logoutput: http://intercity-vpn.de/bootstrap/diff-16-1619253444-and-13-1619350041.html | these are the files
<gforce_de1977> http://intercity-vpn.de/bootstrap/bootstrap.log-multilog-16-1619253444.txt and http://intercity-vpn.de/bootstrap/bootstrap.log-multilog-13-1619350041.txt
<fossy>odd
<fossy>so sometimes it dosent regen man pages
<bauen1>most likely timing issues
<bauen1>but yeah, make --debug is decent for finding timestamp race conditions between multiple runs
<bauen1>there are some differences that are a bit weird: `/after/bin/install -c -m 644 ../../gcc/doc/gpl.7 /after/share/man/man7/gpl.7` vs. `/after/bin/install -c -m 644 doc/gpl.7 /after/share/man/man7/gpl.7`
<gforce_de1977>yeah, that one is really wierd. it is the SAME kernel+initrd, i have to think about it
<gforce_de1977>also i will experiment with overloading the 'date' command, so that we do not have arbitrary timestamps, maybe this changes something and then we can think about it
<bauen1>i'm not sure if SOURCE_DATE_EPOCH (or it was called something like that) is respected by these old programs, but that should make some time issues reproducibility problems disappear
***ChanServ sets mode: +o rekado
<stikonas>gforce_de1977: I doubt that arbitrary timestamps are taken from date
<stikonas>make probably calls time functions from libc directly
<stikonas>you would have to override them in libc
<OriansJ>I guess it is always 31 Dec 1969 forever
<OriansJ>or 1 Jan 1970
<OriansJ>depending on your timezone
<fossy>ok now i can actually say that perl is done
<fossy>now that it works completely
*fossy loves hardcoded /bin paths
<Hagfish>perl done? fantastic
<Hagfish>what future steps does that enable?
<stikonas>Hagfish: it's not necesserily enabling steps
<stikonas>well, newer perl is needed for stuff like automake 1.16 and texinfo
<stikonas>although, these things are optional
<stikonas>I haven't seen anything yet that doesn't build with automake 1.15
<xentrac>there are probably a lot of things from the early 02000s that use Perl in their builds
<stikonas>well, we have perl 5.6.2 with some (not all) modules
<xentrac>fossy: congratulations!
<stikonas>but yes, basically all things use Perl one way or another (especially because of autotools)