Repository navigation
Conversation
a6f934b to
f1ab23a
Compare
|
Salvaged into the fresh re-cut wave: #12932 ( |
|
LM Studio publishes an arm64 AppImage, and omacom/omarchy-pkgs#685 builds |
|
Reviewed at f1ab23a against Defects, both verified:
Competing pull requests:
The second opinion was Codex Medium, working from the three diffs and the package facts above. It found the file-mode defect, which I then reproduced on the worker. It agreed that #11691's availability check is the better basis, and that none of the three is right as it stands. Its independence is not guaranteed. Waiting on the maintainer to choose an approach. The likely shape is #11691's availability check, extended to the rows this pull request covers. |
bin/omarchy-hw-x86-64 was 100644 so git checkouts exited 126 and hid every gated row. Cursor and Zed both ship Linux aarch64 now, so drop those when: gates; keep Edge/Dropbox/Spotify/Ollama/LM Studio.
|
| hw_x86_64() { | ||
| OMARCHY_UNAME_M="$1" "$ROOT/bin/omarchy-hw-x86-64" | ||
| } |
There was a problem hiding this comment.
Real-machine checks are untested
Every test sets OMARCHY_UNAME_M, so none exercises the uname -m path used on real machines. A later mistake in that fallback could hide supported Install rows while all three tests still pass. Add cases with the variable unset and empty, using a stub uname that returns x86_64 or aarch64.
|
Re-reviewed at 257fbf8 against What was checked:
The second opinion was Codex Medium. It found no defect at this head and judged the change proportionate. It agreed that this pull request should carry the fix for #8645, and its independence is not guaranteed. It did contribute two other ways to reach the same failure on aarch64, both verified in source: Setup > Defaults > Browser > Edge ( Other open pull requests on the same rows:
This one is the oldest, and the only one that covers all five rows #8645 reports. The title and description still say eight rows and name Cursor, Zed and Symfony. The head gates five. Waiting on the maintainer, including on the order this and #14431 land in. |
|
Re-checked at 257fbf8 against current What changed underneath it:
Nothing was run this time: the change is the same code verified on 2026-10-07, and what is new is the base around it. These conclusions come from reading The second opinion was Codex Medium. It reached the same breakdown of the five gates, found no other defect at this head, and judged that porting this now would mean choosing how Ollama should behave on ARM, which is not a mechanical merge. Its independence is not guaranteed. Waiting on the maintainer. The recommendation is to close this pull request: #14431 and omacom/omarchy-pkgs#885 cover four of the rows #8645 reports. Ollama on aarch64 is better handled as an aarch64 build in omacom/omarchy-pkgs, or by #11691's fallback, than by hiding the row. If the maintainer would rather hide Ollama for now, the port is small: one |
|
Re-checked at 257fbf8 against current #12932 is not a replacement for this fix on
Nothing was run. These conclusions come from reading the two diffs against Still waiting on the maintainer. The recommendation is unchanged: close this pull request rather than fold it into #12932. #14431 and omacom/omarchy-pkgs#885 already cover four of the five rows #8645 reports. Ollama on aarch64 needs a decision of its own. |
Summary
Gate eight x86_64-only Install-menu rows on
omarchy-hw-x86-64.Fixes #8645.
The catalog offers packages that cannot exist on aarch64 (ollama, lm-studio, cursor, zed, spotify, dropbox, symfony-cli, edge) because the only guard is installed-presence. Add
bin/omarchy-hw-x86-64([[ ${OMARCHY_UNAME_M:-$(uname -m)} == x86_64 ]]) andwhen:gates on those eight rows; presence-disabled stays.Does not hide 1Password (separate tarball installer).
Testing
hw-x86-64-test.sh3 okmenu-test.sh122 ok