Skip to content

bugfix: don't stop early when other spacks are installing#51539

Merged
haampie merged 1 commit intodevelopfrom
bugfix-early-stop-on-multi-node-install
Nov 10, 2025
Merged

bugfix: don't stop early when other spacks are installing#51539
haampie merged 1 commit intodevelopfrom
bugfix-early-stop-on-multi-node-install

Conversation

@tgamblin
Copy link
Copy Markdown
Member

@tgamblin tgamblin commented Nov 8, 2025

Fixes #50933.

The installer in #47590 missed a loop condition and wasn't properly waiting on tasks requeued as BuildStatus.INSTALLING. This means another Spack is working on that task, and we should wait and not bail early.

There is also a condition in the installer where we disocver for the first time that a package is installed in cleanup. We need to update the installed dependencies lists of dependents when that happens.

@rbberger

This comment was marked as resolved.

Fixes #50933.

The installer in #47590 missed a loop condition and wasn't properly waiting on tasks
requeued as `BuildStatus.INSTALLING`. This means another Spack is working on that task,
and we should wait and not bail early.

There is also a condition in the installer where we disocver for the first time that a
package is installed in cleanup. We need to update the installed dependencies lists of
dependents when that happens.

Signed-off-by: Todd Gamblin <[email protected]>
@tgamblin tgamblin force-pushed the bugfix-early-stop-on-multi-node-install branch from 61a768e to 8aa813a Compare November 8, 2025 21:50
Copy link
Copy Markdown
Member

@rbberger rbberger left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Had issues making the current Spack develop fail with my reproducer. However, cherry-picking this commit onto an older commit eaf413e, which still consistently was failing with the reproducer made the failures go away. So, LGTM.

@haampie haampie merged commit 1962a3e into develop Nov 10, 2025
31 of 32 checks passed
@haampie haampie deleted the bugfix-early-stop-on-multi-node-install branch November 10, 2025 07:15
@lowpolyneko
Copy link
Copy Markdown

Can confirm that this PR resolves concurrent runs of spack failing when dependencies are built in parallel. Thanks for the quick fix!

@alalazo alalazo added the v1.1.0 PRs to backport for v1.1.0 label Nov 13, 2025
@alalazo
Copy link
Copy Markdown
Member

alalazo commented Nov 13, 2025

We probably want this bugfix in v1.1.0

@alalazo alalazo mentioned this pull request Nov 13, 2025
17 tasks
alalazo pushed a commit that referenced this pull request Nov 13, 2025
The installer in #47590 missed a loop condition and wasn't properly waiting on tasks
requeued as `BuildStatus.INSTALLING`. This means another Spack process is working
on that task, and we should wait and not bail early.

There is also a condition in the installer where we discover for the first time that a
package is installed in cleanup. We need to update the installed dependencies lists of
dependents when that happens.

Signed-off-by: Todd Gamblin <[email protected]>
alalazo pushed a commit that referenced this pull request Nov 13, 2025
The installer in #47590 missed a loop condition and wasn't properly waiting on tasks
requeued as `BuildStatus.INSTALLING`. This means another Spack process is working
on that task, and we should wait and not bail early.

There is also a condition in the installer where we discover for the first time that a
package is installed in cleanup. We need to update the installed dependencies lists of
dependents when that happens.

Signed-off-by: Todd Gamblin <[email protected]>
Signed-off-by: Massimiliano Culpo <[email protected]>
alalazo pushed a commit that referenced this pull request Nov 13, 2025
The installer in #47590 missed a loop condition and wasn't properly waiting on tasks
requeued as `BuildStatus.INSTALLING`. This means another Spack process is working
on that task, and we should wait and not bail early.

There is also a condition in the installer where we discover for the first time that a
package is installed in cleanup. We need to update the installed dependencies lists of
dependents when that happens.

Signed-off-by: Todd Gamblin <[email protected]>
Signed-off-by: Massimiliano Culpo <[email protected]>
haampie pushed a commit that referenced this pull request Nov 14, 2025
The installer in #47590 missed a loop condition and wasn't properly waiting on tasks
requeued as `BuildStatus.INSTALLING`. This means another Spack process is working
on that task, and we should wait and not bail early.

There is also a condition in the installer where we discover for the first time that a
package is installed in cleanup. We need to update the installed dependencies lists of
dependents when that happens.

Signed-off-by: Todd Gamblin <[email protected]>
Signed-off-by: Massimiliano Culpo <[email protected]>
Signed-off-by: Harmen Stoppels <[email protected]>
becker33 pushed a commit that referenced this pull request Nov 14, 2025
The installer in #47590 missed a loop condition and wasn't properly waiting on tasks
requeued as `BuildStatus.INSTALLING`. This means another Spack process is working
on that task, and we should wait and not bail early.

There is also a condition in the installer where we discover for the first time that a
package is installed in cleanup. We need to update the installed dependencies lists of
dependents when that happens.

Signed-off-by: Todd Gamblin <[email protected]>
Signed-off-by: Massimiliano Culpo <[email protected]>
Signed-off-by: Harmen Stoppels <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

v1.1.0 PRs to backport for v1.1.0

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Parallel builds may fail due to uninstalled dependencies

5 participants