Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-9677

ndarray sometimes fails to build using lsstsw

    XMLWordPrintable

Details

    Description

      Today I asked lsstsw to rebuild a product (astshim) and it failed while bulding ndarray, with this error:

      /Users/rowen/UW/LSST/lsstsw3/build/ndarray/include/ndarray/detail/Core.h:22:10: fatal error: 'ndarray/Vector.h' file not found
      #include "ndarray/Vector.h"
      

      That file is built by m4 so I suspect race condition or other sequencing error in the build instructions. When I resubmitted the rebuild command it worked.

      Attachments

        Issue Links

          Activity

            This happened to me again today. Restarting the rebuild worked (as usual).

            rowen Russell Owen added a comment - This happened to me again today. Restarting the rebuild worked (as usual).
            rowen Russell Owen added a comment -

            And again. Basically it's common when the package needs to be rebuilt.
            I have attached the build log _build.log

            rowen Russell Owen added a comment - And again. Basically it's common when the package needs to be rebuilt. I have attached the build log _build.log
            jbosch Jim Bosch added a comment -

            It's pretty clear where this is happening, but I lack the CMake-fu to (quickly) fix it.  I'm reassigning to pschella in the off chance it is a quick fix for him.

            Ultimately I'd like to get rid of the generated headers entirely, but that will need to wait for some structural changes that I've got on my ndarray to-do list: removing the SCons build, removing Swig support, and switching the Boost.Python bindings to use the Boost.NumPy version now packaged with Boost instead of the submodule (because those also involve generated headers, but the Boost.Python bindings aren't handled at all by the current CMake build).  In the course of doing that I'm sure I'll gain enough of an understanding of the CMake build to fix this problem even if I don't end up dropping the generated headers, so pschella, give this back to me if you don't see an easy fix.

            jbosch Jim Bosch added a comment - It's pretty clear where this is happening, but I lack the CMake-fu to (quickly) fix it.  I'm reassigning to pschella in the off chance it is a quick fix for him. Ultimately I'd like to get rid of the generated headers entirely, but that will need to wait for some structural changes that I've got on my ndarray to-do list: removing the SCons build, removing Swig support, and switching the Boost.Python bindings to use the Boost.NumPy version now packaged with Boost instead of the submodule (because those also involve generated headers, but the Boost.Python bindings aren't handled at all by the current CMake build).  In the course of doing that I'm sure I'll gain enough of an understanding of the CMake build to fix this problem even if I don't end up dropping the generated headers, so pschella , give this back to me if you don't see an easy fix.
            rowen Russell Owen added a comment - - edited

            Almost certainly fixed as part of DM-14728

            If not, re-open this ticket or file a new one.

            rowen Russell Owen added a comment - - edited Almost certainly fixed as part of DM-14728 If not, re-open this ticket or file a new one.
            jbosch Jim Bosch added a comment -

            Looks like this wasn't fixed by DM-14728 after all; price rediscovered it, put together a fix, and I've merged that and cut a release upstream.  I'll use this ticket to push out an LSST uprev of ndarray.

            jbosch Jim Bosch added a comment - Looks like this wasn't fixed by DM-14728 after all; price rediscovered it, put together a fix, and I've merged that and cut a release upstream.  I'll use this ticket to push out an LSST uprev of ndarray.
            jbosch Jim Bosch added a comment -

            price, mind reviewing?  Not much to see in LSST repos, as the only change is the (binary) update to the tarball.  But it got through Jenkins, and ndarray's own travis-ci, and I'm happy to trust your testing that the build race is fixed.

            jbosch Jim Bosch added a comment - price , mind reviewing?  Not much to see in LSST repos, as the only change is the (binary) update to the tarball.  But it got through Jenkins, and ndarray's own travis-ci, and I'm happy to trust your testing that the build race is fixed.
            price Paul Price added a comment -

            Dunno what's really to review, but I approve.

            price Paul Price added a comment - Dunno what's really to review, but I approve.
            rowen Russell Owen added a comment -

            jbosch do you want me to merge this? Have you done it?

            rowen Russell Owen added a comment - jbosch do you want me to merge this? Have you done it?
            jbosch Jim Bosch added a comment -

            Merged.  Thanks for the reminder, and sorry for the delay.

            jbosch Jim Bosch added a comment - Merged.  Thanks for the reminder, and sorry for the delay.

            People

              rowen Russell Owen
              rowen Russell Owen
              Paul Price
              Jim Bosch, Paul Price, Pim Schellart [X] (Inactive), Russell Owen
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Jenkins

                  No builds found.