IRC channel logs

<agg1>strugging with i386-tcc mis-compiled bash.static bug

<agg1>disabled bash-malloc as Mikaku suggested

<agg1>no difference

<agg1>tested several bash versions, bash-2.05, bash-3.2, bash-4.4, bash-5.2

<agg1>it seems bash-2.05 works, but it is insufficient to support bashism inside portage tooling

<agg1>got some gdb backtrace, will reconfirm different bash versions next if it's the same bug with all of them

<agg1>i already got a complete tinycc-compiled distribution here, but it's not self-hosting yet with portage tooling

<agg1>and this is the problem for error-reporting, because a self-hosting distribution release was required

<agg1>so, currently i could throw some emerge.log files and gdb backtraces around, but it's not possible to re-produce this with mentioned tinycc-compiled system

<agg1>i'll try to summarize a report for tinycc mailing list

<agg1>i see bootstrappable compiled bash-2.05 with tinycc, and bash-5.x with gcc

<agg1>what's really strange about this, i got a complete system compiled, linux-2.4/tinycc/musl-1.1, it boots, ssh, nfs, irc, links browser

<agg1>oksh, busybox, all of it works, except bash.static

<agg1>it errors/segfaults on brace expansion, not sure if this is a symptom of any other problem

<agg1>in either tinycc itself or bash-specific

<agg1>it's a nasty show-stopper, because i can't keep portage tooling self-hosting with tinycc due to this

<agg1>i already could hack together a python.static and perl.static which work with portage tooling, compiled/linked with tinycc

<agg1>but i'm pulling my hair out over the bash.static one for a half a year already

<agg1>sigh, found a bash-version which isn't mis-compiled with tinycc AND supports sufficient bashism for portage tooling

<agg1>but it is alot of work remaining to make portage tooling fully self-hosting again on a tinycc-compiled buildhost

<agg1>the bootstrappable build scripting seems appealing, but i fear this one misses cross-compilation support

IRC channel logs

2025-11-17.log