<davidlt[m]>
rwmjones djdelorie nirik I might ask to change kojid.conf once the current builds are finished.
zsun has quit [Ping timeout: 244 seconds]
<rwmjones>
davidlt[m]: any clue about:
<rwmjones>
/usr/bin/ld: .gnu.build.attributes has both ordered [`.gnu.build.attributes.lo.exit' in /lib64/lp64d/../lib64/lp64d/Scrt1.o] and unordered [`.gnu.build.attributes' in /lib64/lp64d/../lib64/lp64d/Scrt1.o] sections
<rwmjones>
there's a very confusing bug report about it for RHEL 9 but I can find nothing about Fedora
<rwmjones>
the ocaml-labltk configure script does:
<rwmjones>
tclversion.c:21:1: warning: return type defaults to 'int' [-Wimplicit-int]
<rwmjones>
21 | main ()
<rwmjones>
| ^~~~
<rwmjones>
Assembler messages:
<rwmjones>
Fatal error: -march=rv64imafdc_zicsr_zifencei: Invalid or unknown z ISA extension: 'zifencei'
<rwmjones>
I'm going to try updating binutils next
<davidlt>
yeah, because RISC-V ISA standard changed (and we used a new default)
<rwmjones>
also
<rwmjones>
error: /usr/lib/rpm/macros.d/macros.build-constraints: line 12: Macro %constrain_build has unterminated body
<davidlt>
the old binutils have don't know that RV64GC means a different thing on a newer spec
<rwmjones>
looks like an actual RPM bug
<davidlt>
these are no longer case IIRC (if you build in Koji)
<davidlt>
i.e. you are missing another package update (don't remember which one)
<rwmjones>
anyway good news is I fixed the ocaml-labltk problem, will push a fix soon
<davidlt>
rwmjones, was that riscv specific or anything terrible?
<rwmjones>
unclear .. it was caused by missing -fPIC when linking the test script, and it seems like it should happen on any arch including x86-64, but I tested it and it only seems to happen on riscv64
<davidlt>
hm..
<davidlt>
redhat-rpm-macros are up-to-date IIRC
<davidlt>
rwmjones, did you check that fPIC is set and passed like on x86_64?
zsun has joined #fedora-riscv
<davidlt>
or are these custom flags used by the test? i.e. ignore what's in redhat-rpm-macros?
<rwmjones>
the configure script is an incredible hack that tries to pull flags out of some ocaml libdir file
<rwmjones>
it gives the same configure problem on x86-64, but for some reason the build still succeeds there
<rwmjones>
the whole thing is junk
<rwmjones>
I just did the min possible patch to add -fPIC
<davidlt>
yeah, that's a small part of reason why I didn't want to look at it ;)
<davidlt>
I assuming it
<davidlt>
it's gonna be something stupid and small, but hunting would be annoying
<davidlt>
actually with more folks working in Koji I might finally go back to sending more patches to upstream and dist-git
<davidlt>
Miro is doing an epic work on Python 3.11. I just now focus on a few broken stuff that's marked as "riscv special"
<rwmjones>
davidlt: shall I copy the fix to ocaml-labltk to f37 branch?
<davidlt>
it's fine to have it in main/rawhide if it's the same as f33
<davidlt>
yeah, I see that main/rawhide/f33 points to the same thing
<somlo>
but, progress! This time I actually got it all the way to the login prompt. I cheated a bit: had a 4-core model before, ran it all (including the dtb specifying 4 cores) on a single-core bitstream
<somlo>
it "failed" to start 3 of the 4 cores it *thought* it had during early boot
<somlo>
so I'm hoping if I redo the dtb to only specify one core I might get it to *not* crash after getting to the login prompt
<davidlt[m]>
somlo: nice!
<somlo>
davidlt[m]: so the earlier crashes (a few days ago) might have something to do with SMP (seeing as trying with a single core got me further). But it's still early in the troubleshooting process :)
zsun has joined #fedora-riscv
zsun has quit [Ping timeout: 244 seconds]
jcajka has quit [Quit: Leaving]
cyberpear has quit [Quit: Connection closed for inactivity]