IRC channel logs

<oriansj>kerravon: well the echo in the reader would effectively echo the source code being built. And if one is paranoid of the stability of one's media: yes plastic and metal strips are options

<oriansj>as there are actual printers one can connect to the echo of the program (line wrapping might take a couple extra instructions though)

<oriansj>sam_: I am glad to hear that the qemu group figured out that exact same bug first.

<oriansj>well the last bugs in M1 that needed fixing are resolved and release 1.5.0 of mescc-tools is now out

<sam_>oriansj: I live to give bug references to people :)

<oriansj>sam_: I live to learn about bugs in my software, so that I can make it even better for everyone ^_^

<sam_>:)

<oriansj>and as a reminder this is a mescc-tools major release and the one we are scheduled to drop support for the flags: --BigEndian, --LittleEndian, --exec_enable and --BaseAddress

<oriansj>and as it has been nearly 2 years and 2 Major releases (with warnings) hopefully everyone converted to the current standard flags (if not: --BigEndian -> --big-endian ; --LittleEndian -> --little-endian ; --exec_enable -> no longer needed ; --BaseAddress -> --base-address )

<stikonas[m]>oriansj: thanks!

<pder>doras: I updated my gcc 13 branch to also disable default pie in addition to ssp. Could you test your build to see if there is an improvement?

<pder> https://github.com/pder/live-bootstrap/tree/gcc-13-disable-ssp

<stikonas[m]>pder: is that ready now?

<stikonas[m]>Maybe make a PR

<pder>From what I can tell, gcc-13 is working fine within the live-bootstrap environment, but doras was running into issues when he moved this toolchain into his system

<pder>I disabled ssp and pie- ssp was causing the segfaults on all binaries. From what I have read enabling default pie can break other stuff including some tests in gcc's test suite

<doras>pder: I haven't tried with `--enable-default-pie` disabled yet, but removing `--enable-default-ssp` definitely solved some of the segmentation faults.

<doras>I can try with this new commit and report.

<doras>This is the commit that I had tested, where most (but not all) of the segmentation faults were solved: https://github.com/pder/live-bootstrap/commit/73ac2eb57c0fe8caad3c22ffe0641cd129b7f7ab

<pder>doras: I am not familiar with how you use the output of live-bootstrap in your build

<pder>do you extact the tarball packages generated onto a different system?

<pder>Also, do you use gcc-13 from live-bootstrap to build a new x64 toolchain?

<doras>pder: I extract the tarballs into an empty directory and effectively chroot (using bwrap) into that with additional source files to build more projects, or the same projects with different configuration (glibc, gawk, file, binutils, gcc).

<doras>pder: the first step is to build a cross-compilation toolchain that uses x86 binaries provided by live-bootstrap (+ a few more we build) and can produce binaries for a few architectures (x86_64, aarch64, ppcle64, riscv64). The next step is to use these toolchains to cross-compile a native toolchain for each architecture.

<doras>We're getting segfaults in the first step already.

<doras>pder: I started CI with the new commit: https://gitlab.com/freedesktop-sdk/freedesktop-sdk-binary-seed/-/jobs/4708661166

<doras>Again, it will mostly be silent for about 7 hours unless a build error comes up (caused by a segmentation fault or otherwise).

<doras>This time I kept the checksum verification enabled.

<pder>doras: I just had a checksum error with musl, so likely your build will also fail

<doras>Yes, it just did.

<pder>musl-1.2.4_2.tar.bz2 specifically

<pder>I will fix that as soon as I can

IRC channel logs

2023-07-21.log