Sentinel: red:01m23vygyffx5hg2f5vm8r74sr is past its budget: trevato/forge: main is red at 0c82bad — "the host holds cells > cells are available" fails; 2 pull checks wait for it t… #6061

This pull stops the check from setting FORGE_REQUIRE_CELLS=1, which should turn main green again. I haven't confirmed that from here. The cell test skips under this sandbox's loopback-port mode, so I couldn't reproduce the failure. I also couldn't run tsc, because bun x tsc --noEmit couldn't find the bun type definitions here and I didn't run bun install. The edit touches only forge.json and a comment. The forge's check will give the real verdict.

Cause. #6031 (0c82bad) added a test, "the host holds cells > cells are available", that throws whenever FORGE_REQUIRE_CELLS=1 and the host can't open a cell. It also set that variable in both the fast and run check commands in forge.json. The check host can't open a cell (my guess is that nested bwrap isn't allowed there), so the test failed on main. That left two pull checks waiting.

Change.

• forge.json: FORGE_REQUIRE_CELLS=1 is removed from the fast and run check commands.
• test/spawn-cell.test.ts: the guard test stays and its comment now says it is opt-in.
• The real-cell test still runs wherever cells work. On a host that sets FORGE_REQUIRE_CELLS=1, an unavailable cell is a failure rather than a skip.

Not done. I didn't try to make cells work on the check host, and I didn't read the reason the probe gave there, because the log isn't visible from here. If the check should require cells, that host needs fixing first and the variable can then go back.

Try. bun test test/spawn-cell.test.ts passes here with 3 skips. The real check is the forge's run on this push.

Closes #6060

This pull stops the check from setting FORGE_REQUIRE_CELLS=1, which should turn main green again. I haven't confirmed that from here. The cell test skips under this sandbox's loopback-port mode, so I couldn't reproduce the failure. I also couldn't run tsc, because bun x tsc --noEmit couldn't find the bun type definitions here and I didn't run bun install. The edit touches only forge.json and a comment. The forge's check will give the real verdict.

Cause. #6031 (0c82bad) added a test, "the host holds cells > cells are available", that throws whenever FORGE_REQUIRE_CELLS=1 and the host can't open a cell. It also set that variable in both the fast and run check commands in forge.json. The check host can't open a cell (my guess is that nested bwrap isn't allowed there), so the test failed on main. That left two pull checks waiting.

Change.

• forge.json: FORGE_REQUIRE_CELLS=1 is removed from the fast and run check commands.
• test/spawn-cell.test.ts: the guard test stays and its comment now says it is opt-in.
• The real-cell test still runs wherever cells work. On a host that sets FORGE_REQUIRE_CELLS=1, an unavailable cell is a failure rather than a skip.

Not done. I didn't try to make cells work on the check host, and I didn't read the reason the probe gave there, because the log isn't visible from here. If the check should require cells, that host needs fixing first and the variable can then go back.

Try. bun test test/spawn-cell.test.ts passes here with 3 skips. The real check is the forge's run on this push.

Main already carries the real fix, b896efa6, which stops the check from setting FORGE_REQUIRE_CELLS=1. The sentinel's "red at 7566693" appears to predate it. I merged main into this branch, and the pull is now one comment edit in test/spawn-cell.test.ts. It says the guard test is opt-in and that the check host can't nest a cell. bun test test/spawn-cell.test.ts passes here (3 skipped, as the sandbox can't make cells). I haven't confirmed main is green.

Left out: I didn't make cells work on the check host, so the "cells are available" guard only fires where FORGE_REQUIRE_CELLS=1 is set by hand.

To close the issue: if the forge's check on main at or after b896efa6 is green, the sentinel's red cause should clear and the issue can be closed without merging this. If the check is still red there, its log will say why.