- Status Closed
- Percent Complete
- Task Type Bug Report
- Category Packages
-
Assigned To
Andreas Baumann - Operating System i686
- Severity High
- Priority Very Low
- Reported Version
- Due in Version Undecided
-
Due Date
Undecided
- Votes
- Private
Attached to Project: Arch Linux 32
Opened by Piotr Cieśliński - 19.03.2023
Last edited by Andreas Baumann - 07.04.2023
Opened by Piotr Cieśliński - 19.03.2023
Last edited by Andreas Baumann - 07.04.2023
FS#329 - [go] Cannot install yay from AUR
Installing yay from AUR
git clone https://aur.archlinux.org/yay.git cd yay makepkg -sri
gives following error
go: updates to go.sum needed, disabled by -mod=readonly
Reason - outdated version of go in i686 architecture
Temporary solution - use go for pentium4 (change temporarily architecture to pentium4 in /etc/pacman.conf)
Closed by Andreas Baumann
07.04.2023 06:00
Reason for closing: Fixed
Additional comments about closing:
07.04.2023 06:00
Reason for closing: Fixed
Additional comments about closing:
bootstrapped for i686 and pentium4, i486
not available anymore..
Your temporary solution is not at all valid because pentium4 depends on sse2 and go will install but it wont start… its something linked with sse
Yep, correct. So. yay is not for i686 then..
@abaumann
https://go-review.googlesource.com/c/go/+/258957
these commits might be helpful..
after doing a couple of patching an i686 go can be compiled…
Thanks for the link. The patch looks like enabling i87 over SSE math again, which is handy,
will test.
But if I remember correctly there was also a SSE2-thingy lurking somewhere.
https://go-review.googlesource.com/c/go/+/257963
after patching them linux-386-387 target can be comeback…
@abaumann
are there any signs of a working go on i686?
So, just changing buildall.bash is not enough, one has to backport all of the x87
support back to 1.20.2.
I would rather opt to use a different go compiler in this case (like the GNU one).
ok gcc-go cannot compile yay. I would suggest to switch to some other AUR
helper.
wait isnt it possible that after patching these.. https://go-review.googlesource.com/c/go/+/257963
lemme check others…
i686 is too old, so I would have to bootstrap go from 1.14 to 1.17 (in Rust style, no thanks):
Tests are failing, which is usually a bad indicator for a working compiler..
yep, i486 works. now I have to see how I can bootstrap go on i686.
On pentium4 everything is fine.
The buildmaster is in maintainance currently, so there will be a while till the
builds are picked up..
1.17 doesn't understand softfloat, great..
..ok, back to go15, then try to bootstrap from there.
@abaumann
Are there any signs of luck? and where is 486 go?
486 go was just working on a i486 installed Alix Geode (which has MMX support).
src/go/src/runtime/asm_386.s has a nice CPU check which emits:
"This program can only be run on processors with MMX support."
on a real 486 (or on a 486 VM). So i686 and pentium4 might work after
bootstrapping, but I have my doubts about 486 can be supported at all.
https://github.com/ish-app/ish/issues/57 claims there is only a movq which
is used for atomic operations.. this might be implemented for i486.. but
first I would like to have it working on i686 and pentium4..