garlick commented on issue flux-framework/flux-core#7494.
Reproduced:…
vsoch open issue converged-computing/mcp-server#10.
Features to add for more production robustness
When we have this deployed in a production setting, I want:…View Comment
markcmiller86 commented on issue visit-dav/visit#20892.
Ok, I built noise.C on current develop against Silo 4.12 and Silo 4.10.3 and then ran both versions of noise creating PDB files….
cyrush commented on issue visit-dav/visit#20892.
I think maybe its an issue with looping over the dims…
imguoguo commented on issue sipeed/picoclaw_docs#18.
LGTM…
v-dobrev commented on issue mfem/mfem#4917.
Merged in next for testing……
alecbcs commented on issue spack/spack#52144.
> I think putting new commands we just happen to want in CI under spack ci is not really the path forward. This is going to operate on mirrors, considering the state of a repo. Neither of those things has anything to do with “ci” which is really just “build farm scheduling”. …
cjvogl commented on issue mfem/mfem#5270.
I just ran the code and everything is looking good! All that is left is to address the last syntax concerns noted above….
artv3 commented on issue mfem/mfem#5124.
Any comparisons with MFEM’s AMGX backend? https://github.com/mfem/mfem/blob/master/examples/amgx/ex1p.cpp Super cool! …
mplegendre commented on issue llnl/Spindle#148.
I went over the /bin/.. case in detail, and it looks like we handle it okay. There’s definitely some oddities with spindle putting “..” into an internal variable named ‘file’. But we ultimately always supported spindle operations on directories (even if that wasn’t common), and “..” just gets treated like other directories and handled correctly. …
markcmiller86 commented on issue llnl/Silo#527.
@biagas this is how I’d like to proceed with Silo. That said, if you feel better about your original patches for bv_silo.sh because you have tested those changes in multiple places, then I am ok with keeping your bv_silo.sh logic as is….
garlick commented on issue flux-framework/rfc#505.
Eh this was prematurely posted - sorry. Closing for now….
jameshcorbett commented on issue flux-framework/flux-sched#1432.
Aside from 229e25d, there are no code changes from @milroy ‘s last review. However, I did add some new tests in the last commit. The tests fail because the agfilter output is incorrect for cores: it considers fewer cores to be allocated than are in fact allocated. This discrepancy appears to have something to do with the match format and the way it skips vertices, because if the test is re-run with rv1 match format instead of rv1_shorthand, or if the job is modified to only request nodes non-exclusively, the agfilter checks pass. …
jameshcorbett open issue flux-framework/flux-core#7494.
idset: `idset_set (idset, -586)` caused `SIGABRT`
As part of https://github.com/flux-framework/flux-sched/pull/1438 I tried passing some negative integers to idset_set and found that passing -586 (which I picked randomly) caused my process to be killed with SIGABRT. I noticed the following assertion message in my test output:…View Comment
grondo commented on issue flux-framework/flux-core#7495.
I’d just add a note in the documentation. Something like…
wihobbs commented on issue flux-framework/flux-core#7495.
I guess an edge case might be “what if the hostfile for the taskmap specifies fewer hosts than the -N request?” which is currently allowed….
jameshcorbett commented on issue flux-framework/flux-coral2#463.
Similar to your PR this also needs a manual merge @cmoussa1 …
cmoussa1 commented on issue flux-framework/flux-accounting#836.
Thanks @jameshcorbett! Setting MWP here…
JustinPrivitera closed issue visit-dav/visit#20846.
numpy issue on LC toss4
Even though default compiler on lc systems is gcc 10, numpy will look for ccache, which in a default env ends up using gcc 8.x. …View Comment
becker33 closed issue spack/spack#51971.
Mirroring packages that throw an error breaks mirroring
### Steps to reproduce…View Comment
hppritcha commented on issue open-mpi/ompi#13790.
okay this is curious. The “problem” seems to be in the ompi error checking functions in cases where the send/recv count value is 0. This is not really a fortran interface problem….
tzanio commented on issue mfem/mfem#5265.
This PR is now under review (see the table in the PR description). To help with the review process, please do not force push to the branch….
cyrush open issue llnl/conduit#1557.
python DataType constructor should accept same args as set
```…View Comment
cyrush commented on issue llnl/blt#758.
Sounds like: https://cmake.org/cmake/help/latest/prop_tgt/CXX_STANDARD_REQUIRED.html#prop_tgt:CXX_STANDARD_REQUIRED…
artv3 closed issue llnl/RAJAPerf#666.
MASS3DPA_ATOMIC checksum failure
Describe the bug…View Comment
trws commented on issue flux-framework/flux-sched#1438.
You could write a catch2 test that literally calls the function. That’s probably what I would do actually, there’s no reason to make it an integration test….
chu11 commented on issue flux-framework/flux-core#7136.
was chatting with @watson6282 about general “how to know about a certain failure” was brought up. I recalled this issue….
cbritopacheco commented on issue cbritopacheco/rodin#204.
@copilot+claude-opus-4.6 Please now implement the generic src/Rodin/Solid/Linear/LinearElasticityIntegral.h for any FES. Take a look into the P1 specialization for LinearElasticityIntegral and H1 QuadratureRule implementations for the general idea….
jedbrown commented on issue CEED/libCEED#1808.
In the makefile, we figure that discerning users should set OPT themselves. Our default is roughly like CMake’s RelWithDebInfo, though I think that uses -O2 by default. I would not be opposed to switching to making the default -O2 (it’s basically a tradeoff of debug sensibility, but you likely still want more aggressive (e.g., associative math) for production runs. (The significance of such flags depends on how the user’s qfunctions are written; the library parts of libceed are not so sensitive.)…
cyrush open issue Alpine-DAV/ascent#1681.
rendering at resolution greater than 1024 is broken because of bad tiled rendering
I have found a regression running Ascent with a single task, but at resolution greater than 1024…View Comment
jeremylt commented on issue openjournals/joss-reviews#9153.
Realistically, I’m probably jammed up until August, unfortunately…
balos1 commented on issue llnl/sundials#809.
> I think that might be ok (as long at the results are not radially different). The runner images are not updated frequently, so this would be a way of moving to newer versions as the “latest” runner is updated. …
becker33 closed issue llnl/llnl.github.io#734.
Repository owners not included in Search context
When running a search in our catalog, I would expect to be able to search for a specific GitHub organization to find all of their entries, especially since the cards are presented with the “owner/name” format for the titles. However, as an example, searching for “chaos” only returns chaos/chaos.github.com….View Comment
JustinPrivitera commented on issue llnl/conduit#1362.
extrapolating from this limited data:…
grondo open issue flux-framework/flux-sched#1437.
invalid idset is output in R
Over in flux-framework/flux-core#7486 a change in configuration resulted in Fluxion generating R with invalid idset in the core field:…View Comment
benzwick open issue benzwick/StreamVorti#39.
Fix the sparsity pattern plots
The zoomed out ones look wrong….View Comment
jedbrown commented on issue RDycore/RDycore#381.
Spelunking the history: …
jeremylt commented on issue CEED/libCEED#1808.
Ok, I think I’m following now….
trws commented on issue spack/spack-packages#3962.
I’m generally supportive of the change, but is there a mechanism we can use to make this easier on users? If the ruff and ty builds work the same as the py- versions for all practical purposes, can we keep the py- names as aliases for example?…
tgamblin commented on issue spack/spack#52061.
> it’s simpler to have fewer exceptions to the rule. …
alecbcs open issue llnl/hubcast#242.
Permissions sync between forges
Repository owners will need to define their maintainers and permissions structure on both the source and destination forges. It may be cumbersome to maintain both of these permissions list simultaneously….View Comment
BradWhitlock closed issue llnl/axom#1822.
Build problem with -DAXOM_NO_INT64_T definition
When AXOM_NO_INT64_T is defined, some mpi_traits in axom/core/Types.hpp are not defined. This later causes a compilation problem in quest where some MPI reductions rely on int64 support….View Comment
alecbcs commented on issue spack/spack-packages#3906.
It looks like version.satisfies() doesn’t operate like I’d hoped. I’ve asked @LinaMuryanto to revert the previous two commits for now and I’ll work on making an issues / fix upstream to make this better for future PRs….
btalamini open issue mfem/mfem#5281.
Silent enzyme errors
# Summary…View Comment
btalamini open issue EnzymeAD/Enzyme#2765.
Promote Enzyme pass remarks to warnings when they indicate result may be incorrect
I have a case in which a derivative is silently computed incorrectly. If I compile my code with the flag -Rpass=enzyme, the following remark is generated:…View Comment
tgamblin commented on issue spack/spack-packages#3945.
Looks like py-kornia-rs and friends are building now and CI is fixed….
cbritopacheco commented on issue cbritopacheco/rodin#201.
@copilot+claude-opus-4.6 I have made some fundamental changes to polytope iteration in the library. Namely, I have changed that MPI Mesh polytope iterators return with the MPI Mesh reference, as opposed to before that they returned the shard’s reference. This has a lot of implications in the library. The first one is that XDMF somehow encounters this exception. Please fix it and find other places in the library which should be modified. …
alecbcs commented on issue spack/spack-packages#3931.
Should this package also depend on openjpeg itself?…
tgamblin commented on issue spack/spack#51931.
@johnwparent take a look at the simplifications here, particularly around locking vs. atomic move….
vsoch commented on issue kubeflow/trainer#3325.
@andreyvelich this is ready for review and eventual merge - apologies for my confusing posts! I was testing locally with the wrong container image (I had updated it back to kubeflow for the PR here, and needed to use my custom build). …
gauthier12 closed issue jetpax/pycoclaw#3.
Source code
Hi, …View Comment
grondo open issue flux-framework/flux-sched#939.
resources are not scheduled fairly among competing queues of job requests
Problem: when multiple queues are defined, fluxion schedules all jobs in the first queue before looking at the next queue. The queues are thus implicitly prioritized according to the order defined, with no available mechanisms to prevent starvation or ensure progress for all queues. …View Comment
grondo open issue flux-framework/flux-core#4306.
use cases for partition + qos or queues
Here are the basic things that we do with the combination of partitions and qos in Slurm, and just queues in LSF: …View Comment
garlick commented on issue flux-framework/flux-core#4306.
That is almost exactly what was kicking around in my brain. I was internally calling them “vqueues” but external sounds good too….
vsoch commented on issue flux-framework/flux-core#7217.
Go home, codecov, you’re drunk….
vsoch commented on issue flux-framework/Tutorials#57.
This came out great! Since our HPSF tutorial was shorter than anticipated, I recorded the entire thing (AWS, cpu with EFA and GPU) today with a short (2.5 minute) introduction to Flux. It is edited and ready to go, and will be released with the Kubeflow blog announcement….
benzwick commented on issue benzwick/SlicerAdaptiveBrush#4.
Hi @Thanhhuynh9 I tested this on my MacBook and it seems to work now. I will close this issue but please reopen it if you still have problems with installing this extension….
cyrush open issue visit-dav/visit#20871.
anari volume plot issue when general anari rendering support is enabled
### Describe the bug…View Comment
becker33 closed issue spack/spack#45173.
Allow arbitrary packages in module hierarchy
### Summary…View Comment
garlick commented on issue pmodels/mpich#7751.
BTW our two projects are in the Linux Foundation together now. IMHO we should try hard to make Flux and MPICH work great together!…
tzanio commented on issue mfem/mfem#5269.
This PR is now under review (see the table in the PR description). To help with the review process, please do not force push to the branch….
gardner48 commented on issue llnl/sundials#880.
@mottelet was this a duplicate of #753? …
BradWhitlock closed issue llnl/axom#1780.
Resolve conduit memory handling
We currently have 2 ways to use Axom memory in Conduit. They were developed as independent utilities in different efforts. This issue is for discussing how to combine them to remove redundancies….View Comment
chu11 commented on issue flux-framework/flux-core#7448.
> If there are multiple suggestions they should be deduplicated. I’m not sure this will happen in practice, flux-python is installed to /usr/bin so won’t appear as a discovered command in a normal installation….
trws commented on issue flux-framework/flux-core#7451.
I just finally got time to actually read this over. This is a massively more complicated scheduler setup than I expected from the “toy” designation. It’s a very nice addition, especially as a way to let people experiment with new custom schedulers. I’m a bit tempted to make a few adjustments to this to make it work more the way qmanager does from a progress and behavior perspective, it might actually not be all that much of a lift to replace qmanager with a reworked version of the backfill class here. We’d have to do some performance tests and similar, but I’m sorely tempted to give it a shot, especially given some of the features we’ve been discussing for that recently the flexibility would be nice, and it isn’t a compute-bound service either. Do you think we could find some time to chat over this next week @garlick? …
cmoussa1 open issue flux-framework/flux-accounting#829.
idea: emulate fair-share calculation for a given association or hierarchy
While presenting at HPSF this week, I got an interesting question from an attendee about how often we tinker with our various priority factors if we don’t like how job priorities are being calculated. While the answer was pretty straightforward (meaning that we don’t adjust the weights for these factors very often), it did get me thinking about something sort of related… I wonder if it would be useful to be able to simulate how fair-share would be calculated for a hierarchy of banks and associations. Something like loading a sample YAML or JSON file of the relevant association and bank information needed to calculate fair-share and just outputting what the fair-share values would be for the hierarchy….View Comment
jeremylt commented on issue CEED/libCEED#1944.
Ok, I have largely fixed the worst oddities, I think…
cyrush commented on issue Alpine-DAV/ascent#1669.
@aroccaPrediqtt …
JustinPrivitera closed issue visit-dav/visit#20777.
Audit host profiles for release - which ones are out of date?
https://github.com/visit-dav/visit/tree/develop/src/resources/hosts…View Comment
JustinPrivitera closed issue llnl/conduit#1497.
Relax material ordering restriction for material dependent field, specset, and matset style converters
Material dependent field style converters need not present materials in the same order as their associated matset. The same is true of specsets. Yet there is an artificial restriction imposed by the material dependent field and specset style converters….View Comment
jameshcorbett open issue flux-framework/flux-coral2#458.
Rabbits: use a new property to exclusively allocate rabbits
In an offline discussion between @behlendorf @garlick @kkier and Py, we were brainstorming ways to hand the rabbits over to users on El Cap. The primary problem is that El Cap does not have JGF enabled, since slices of JGF stored repeatedly in the KVS for each job will occupy too much storage and memory (see https://github.com/flux-framework/flux-core/discussions/7464 for details)….View Comment
JustinPrivitera commented on issue Alpine-DAV/ascent#1675.
I wrote a converter from polyhedral mixed to mixed (?) in VisIt; perhaps that would be of use here?…
emily-howell commented on issue Alpine-DAV/ascent#1494.
I am having trouble reproducing this bug. There is already a test (https://github.com/Alpine-DAV/ascent/blob/74d78e80f839eb1cace5a48e85aed2696729f584/src/tests/ascent/t_ascent_triggers.cpp#L488) that generally reproduces this .yaml and is passing. Even after modifying it to be an exact match to this .yaml example I was unable to reproduce the bug….
markcmiller86 commented on issue visit-dav/visit#20860.
@JustinPrivitera and @visit-dav/visit-core-team I have now tried to handle hyphenations. …
v-dobrev commented on issue spack/spack-packages#1469.
It looks like CI now requires SHA256 checksum for packages – should be an easy fix. …
v-dobrev open issue mfem/mfem#5275.
cudaErrorInvalidDevice abort in GPU builds due to missing device migration in HypreParMatrix::DropSmallEntries
When running a GPU-accelerated build (CUDA + Thrust), calling HypreParMatrix::DropSmallEntries results in a poisoned CUDA context that causes subsequent device operations to fail with a hardware abort. …View Comment
balos1 closed issue llnl/sundials#844.
Bug in SUNLogger_SetXFilename?
Hi all,…View Comment
adrienbernede commented on issue llnl/radiuss-spack-configs#162.
Interesting issue. So, if I understand correctly, the concretization issue comes from us enforcing a compiler dependency on a package that has none. But when removing the compiler, we lack a constraint to enforce coherency between the compiler / software stack version and the external package version… …
markcmiller86 open issue llnl/Silo#520.
Fix funky line endings
Some of the source files in fpzip and maybe elsewhere have funky line endings. These should be fixed….View Comment
adayton1 commented on issue hypre-space/hypre#1490.
We’ll give it a try!…
jeremylt commented on issue 4C-multiphysics/4C#1870.
Ah, that makes sense. I’d then personally try to create a way to bypass those safeguards and guard that header with warnings…
artv3 commented on issue llnl/RAJAPerf-Benchmark#4.
Looks good to me! …
adrienbernede commented on issue spack/spack-packages#3804.
Changes look good to me. Concretization and build tested internally here: …
trws commented on issue flux-framework/flux-accounting#793.
There are a few more considerations here that hit a couple of different use-cases….
grondo commented on issue flux-framework/flux-core#7453.
> The flux start –test-conf idea is a good one on its own IMHO although certainly scoped wider than the immediate need here and may get a bit tricky in combination with other options and environment variables, whereas flux alloc is a bit more constrained….
jedbrown open issue cu-numpde/spring26#1.
Sign up for community presentations
Please comment here with the name of the software (linked to the website), and a sentence or two about what it does. Please make your repository public [1] and link your proposal.md below. (You are encouraged to comment here as soon as you’ve chosen a package; you can keep working on the proposal.md file.)…View Comment
tomstitt closed issue openai/codex#14672.
bwrap sandbox bind fails with symlinked `TMPDIR`
### What version of Codex CLI is running?…View Comment
adrienbernede commented on issue llnl/radiuss-shared-ci#73.
> You might check if you can build locally– I see a couple of minor formatting issues if I do sphinx-autobuild docs/sphinx docs/_build/html
…
BradWhitlock commented on issue llnl/axom#1780.
I finally ran into the problem that Brian reported when I could not make Bump and MIR use a custom allocator for a memory pool. I am adopting Brian’s approach….
garlick commented on issue flux-framework/flux-core#7453.
Although, on broker attributes, we could do something like add modprobe.* or module.* to the broker attribute list and adopt that as a standard way of selecting alternatives. Less bad that having to add a new one for every new alternative….
cbritopacheco commented on issue cbritopacheco/rodin#188.
@copilot+gpt-5.3-codex I have re-organized the structure. I would like you to: …
JustinPrivitera open issue visit-dav/visit#20862.
Blueprint plugin - give material map to matsets before calling to_silo
Say we have a case like this…View Comment
cyrush commented on issue visit-dav/visit#20857.
Note: LC may remove the broken PySide2 module from their python 3.13 install….
jeremylt commented on issue openjournals/joss-reviews#9153.
This looks interesting and I’m happy to review, but it may take me a bit longer than usual, as I’m about to do a big move. If the timeline is shorter then I’m likely not the best option….
tzanio commented on issue mfem/mfem#5252.
This PR is now under review (see the table in the PR description). To help with the review process, please do not force push to the branch….
adrienbernede commented on issue llnl/Umpire#1056.
@davidbeckingsale This is ready!…
adrienbernede commented on issue llnl/RAJA#1999.
@rhornung67 This is ready!…
adrienbernede commented on issue llnl/Caliper#727.
@daboehme This is ready!…
adrienbernede commented on issue llnl/CHAI#360.
@adayton1 This is ready!…
vsoch open issue kubeflow/trainer#3332.
[feat] Feature requests for Flux Policy
### What you would like to be added?…View Comment