← Back to issue list

Missing items in LD_LIBRARY_PATH break numpy stage-packages

View original Launchpad issue

Metadata

Project
snapcraft (launchpad)
Number
#1862516
Type
issue
State
open
Author
~aglavic
Labels
—
Created
Updated
Closed
—

Current evaluation

Numpy staged via stage-packages fails at runtime because libblas.so.3 and liblapack.so.3 sit in nested blas/lapack directories outside the library search path; the old issue is untriaged.

Suggested action: needs triage

Reason: The report identifies exact missing libraries and nested directories, supplies a concrete setup, and shows an LD_LIBRARY_PATH workaround, so a packaging-path defect is plausible. However, it has no labels, maintainer comments, or current reproduction; current dependency documentation still requires users to add missing library directories, so the behavior needs verification and prioritization rather than closure.

Evaluated by: mr-cal-bot

Impact: 55 Quick Win: 16 Complexity: 58 Confidence: 72 Actionability: 70

Issue body

When trying to build a python project that used numpy by adding python3-numpy to stage-packages the import fails because libblas.so.3 and liblapack.so.3 cannot be found. The libraries are included in the snap package bud their path isn't searched during execution. I could work around it by adding these lines to my snapcraft.yaml file:     environment:       LD_LIBRARY_PATH: $LD_LIBRARY_PATH:$SNAP/usr/lib/x86_64-linux-gnu/blas:$SNAP/usr/lib/x86_64-linux-gnu/lapack My setup is: - VirtualBox running Ubuntu Eoan (x86_64) - Latest snapcraft from snap - Using lxd based build I assume the issue is that these libraries use a sub-directory and thus this could be a general issue in the packaging process of libraries. The issue doesn't come up when using numpy from python-packages, but this is no option when using complex dependencies like wxpython that need binary packages and depend on numpy.

Evaluation history

Date Model Scores Action Summary
stealth/space-bunny-alpha
Impact: 55
Quick Win: 16
Complexity: 58
Confidence: 72
Actionability: 70
needs triage Numpy staged via stage-packages fails at runtime because libblas.so.3 and liblapack.so.3 sit in nested blas/lapack directories outside the library search path; the old issue is untriaged.
stealth/space-bunny-alpha
Impact: 48
Quick Win: 22
Complexity: 35
Confidence: 78
Actionability: 72
needs triage A 2019 report says python3-numpy from stage-packages fails because BLAS/LAPACK in architecture-specific subdirectories are absent from LD_LIBRARY_PATH. No maintainer response or reproducer was recorded; current docs prescribe configuring the app's LD_LIBRARY_PATH.
qwen/qwen3.8-27b
Impact: 20
Quick Win: 12
Staleness: 95
Complexity: 40
Confidence: 70
Support Request: 55
close stale 2019 report that stage-packages python3-numpy fails at runtime because libblas/liblapack live in subdirectories not on the library search path; reporter found an LD_LIBRARY_PATH workaround. No labels, comments, or maintainer activity in 6.5 years.
qwen/qwen3.6-35b-a3b
Staleness: 98
Complexity: 35
Confidence: 85
Support Request: 20
close stale Build failure when using python3-numpy in stage-packages due to missing BLAS/LAPACK libraries in LD_LIBRARY_PATH. Unresolved for over 6 years with no maintainer response.
qwen/qwen3.6-35b-a3b
Staleness: 98
Complexity: 45
Confidence: 95
Support Request: 60
close stale Reports numpy stage-packages failing at runtime due to missing BLAS/LAPACK library paths in snapcraft. Provides a workaround. Unaddressed for over 6 years on an EOL Ubuntu release.
qwen3.6-35b-a3b-mtp-q6
Staleness: 95
Complexity: 35
Confidence: 85
Support Request: 25
close stale Runtime failure when importing numpy from stage-packages due to unresolved BLAS/LAPACK library paths. Issue has zero comments, no labels, and has been inactive for over six years.

Update history

Date Change
created

Related work

  • Related To: snapcraft#462 (confidence 65%)

    Historical commit a983e470b added SNAP_LIBRARY_PATH to LD_LIBRARY_PATH, providing related context but no evidence that it fixed nested blas/lapack paths.

Related issues

No related issues found above the similarity threshold.