Conversation
|
Thank you for the patch! The ucx/doc changes look fine to me. Regarding the download error that you see, it seems that there is something wrong in the TLS negotiation. Can you post the output of the following command here?: |
|
As requested : |
|
Thanks, it seems that your curl requires TLS v1.3, while our server only provides TLSv1.2. Could you check whether specifying a TLS version (like here: https://superuser.com/questions/606598/specifying-minor-tls-version-when-using-curl) works around this? EDIT: Actually, the server only seems to supports TLSv1.0. |
|
Doesn't seem to help : |
|
Yeah, I think it only supports TLSv1.0. |
|
That didn't help either : I'm on centOS-8.1 : |
|
Is there a reason to not switch the url to github repo ? |
Currently, the archive on the GitHub repo is autogenerated and not the same archive that we provide on our server. The difference between the archives is that GitHub one requires autoconf/automake, while ours doesn't. I'll discuss with the group regarding what we can do to resolve this. |
|
The current spack recipe places the build in the spack defined build directory, but the default charmpp behavior is to place it in the I'm asking this because I tried compiling |
I switched the package to download via http instead of https until we update our server (#15667). |
|
Merge #15666 instead. |
As per charmplusplus/charm#2477, an extra configuration step is needed when
backend=ucx, hence I added a variantucx_pmi. The default setting ofucx_pmidoes not break the current logic but I'm not sure if there is a better way to do this (which would require a variant that is defined only if another variant is set to a particular value! ).The doc error no longer exists for the latest charmpp release, hence it should only be fixed for older versions.
Fetching the latest version fails on develop, thus I switched to the github url but the checksums don't match. Perhaps someone who knows more about this package can point out what's going on.
Requesting @matthiasdiener for review.