Conversation
31aa625 to
f8c53cc
Compare
vicentebolea
left a comment
There was a problem hiding this comment.
ParaView bites looks great
f8c53cc to
7f2ba1a
Compare
vicentebolea
left a comment
There was a problem hiding this comment.
libxslt is failing in the CI
|
Added remaining missing maintainer. |
|
FYI. |
8f25dff to
3de6adc
Compare
cbc67c7 to
c9d95d4
Compare
7b8cd9a to
a434755
Compare
e173af6 to
bea28fe
Compare
|
@kwryankrattiger I can't help but note that this PR is quite old and we are still waiting for ParaView 6 to be available in Spack. What is the status of this? |
|
There are a couple of files with conflicts that need to be resolved. |
Yes, it is quite old :( The main blocker on this was that an unrelated package was getting stuck on a concretization error. To address this more systematically the X, GL, and Qt updates were extracted into separate PRs. These two PRs that have since found themselves hung up in review:
These are being ignored, there will be a rebase to only contain the paraview related changes at some point. I am hopeful to remaining two PRs will get merged sooner rather than later. |
* Updates to X related packages spack#192 * fix typo in options Co-authored-by: Wouter Deconinck <wdconinc@gmail.com> --------- Co-authored-by: Wouter Deconinck <wdconinc@gmail.com>
7c6875e to
446c255
Compare
author Ryan Krattiger <80296582+kwryankrattiger@users.noreply.github.com> 1771535312 -0600 committer Ryan Krattiger <ryan.krattiger@kitware.com> 1771541345 -0600 parent b411447 author Ryan Krattiger <80296582+kwryankrattiger@users.noreply.github.com> 1771535312 -0600 committer Ryan Krattiger <ryan.krattiger@kitware.com> 1771541285 -0600 Updates to Qt6 related packages spack#192 (spack#2058) * Updates to better support Qt6 for related packages spack#192
* VTKm -> Viskores transition * Drop `use_vtkm` variant for 6: (use variants rocm/cuda/fides) * External Viskores * Dynamic loading GL with GLAD * Qt6 support * Variant for X11 (previously implicit with GLX)
446c255 to
028a6fc
Compare
It should be allowable to run paraview against a GL not in the DAG.
This is not a full fix to this issue. It is still possible to construct cases where an externally built glib-bootstrap is not detected correctly, or a package that depends on glib and glib-bootstrap will always find the bootstrap version of glib-2.0. In order to handle those cases correctly more thought is needed in how this should be expressed via the package API/spec.
Missing libpxbackend.
Newer versions (head of master) works with Qt >= 6.10. This should be bumped or dropped for the next release.
97bfcee to
37b878c
Compare
| # TODO: Fix this in a more complete way | ||
| # This needs to be manually added to the PKG_CONFIG_PATH or it won't be detected by | ||
| # dependent packages reliably. | ||
| # If bootstrap is external append it to ensure that Spack paths are searched before | ||
| # system paths | ||
| # If it is built by Spack it should be safe to prepend. | ||
| if self.spec.satisfies("^glib-bootstrap"): | ||
| env.append_path("PKG_CONFIG_PATH", self.spec["glib-bootstrap"].prefix.lib.pkgconfig) | ||
| glib_bootstrap_spec = self.spec["glib-bootstrap"] | ||
| pkg_config_path = self.spec["glib-bootstrap"].prefix.lib.pkgconfig | ||
| if glib_bootstrap_spec.external: | ||
| env.append_path("PKG_CONFIG_PATH", pkg_config_path) | ||
| else: | ||
| env.prepend_path("PKG_CONFIG_PATH", pkg_config_path) |
There was a problem hiding this comment.
The root of the issue is the link dep of a build dep is not included in the PKG_CONFIG_PATH variable. It is labeled, oddly, as a runtime only dep. To fix that is going to require a lot more thought as we have to change logic that is buried inside of build environment and my first pass at that was incomplete. This is easier and fixes the most immediate case.
You can still have a case where glib and glib-bootstrap are both in the deps and the glib-2.0 is always detected as the bootstrap version even if that isn't what is wanted.
--
So on my system I have an external package configured also in this stack. This creates a problem with pkg-config search for glib-2.0 finds the wrong glib.
When we build the search path for spack we prepend all externals and then all non externals. This way we search the spack built prefixes first, then external prefixes. The external prefix that I have is not excluded and also contains the pkg-config files for all of my system packages. Including an external glib that doesn't satisfy the version constraint for gobject and is the wrong glib I'm general. So appending this package path always results in broken search in a common case.
So to "fix" this case I try to mimic the behavior of the build environment where spack installed packages are listed first, and externals are listed second. This is not a complete fix, but it gets past this common case for now.
Updates for ParaView 6