Odd ordering of GitHub Releases due to numerical sort?

[dnwe]

I happened to notice that the GitHub Releases of the various binaries seem to be showing up in a strange order

e.g., for semeru17-binaries jdk-17.0.10+7_openj9-0.43.0 is the newest and latest release but it doesn't show up until page 2 of the https://github.com/ibmruntimes/semeru17-binaries/releases page, instead the older releases show up there.

I'm guessing this is related to the naming of the releases and because GH is incorrectly doing some sort of numerical sort (sort -nr) whereby 17.0.10 comes after 17.0.2

jdk-17.0.9+9_openj9-0.41.0   
jdk-17.0.8.1+1_openj9-0.40.0 
jdk-17.0.8+7_openj9-0.40.0   
jdk-17.0.7+7_openj9-0.38.0   
jdk-17.0.6+10_openj9-0.36.0  
jdk-17.0.5+8_openj9-0.35.0   
jdk-17.0.4.1+1_openj9-0.33.1 
jdk-17.0.4+8_openj9-0.33.0   
jdk-17.0.3+7_openj9-0.32.0   
jdk-17.0.2+8_openj9-0.30.0   
jdk-17.0.10+7_openj9-0.43.0  
jdk-17.0.1+12_openj9-0.29.1  

The CLI gives a more sensible view (sorted by date by default):

% gh release list --exclude-pre-releases
TITLE                         TYPE    TAG NAME                      PUBLISHED
jdk-17.0.10+7_openj9-0.43.0   Latest  jdk-17.0.10+7_openj9-0.43.0   about 22 days ago
jdk-17.0.9+9_openj9-0.41.0            jdk-17.0.9+9_openj9-0.41.0    about 3 months ago
jdk-17.0.8.1+1_openj9-0.40.0          jdk-17.0.8.1+1_openj9-0.40.0  about 6 months ago
jdk-17.0.8+7_openj9-0.40.0            jdk-17.0.8+7_openj9-0.40.0    about 6 months ago
jdk-17.0.7+7_openj9-0.38.0            jdk-17.0.7+7_openj9-0.38.0    about 9 months ago
jdk-17.0.6+10_openj9-0.36.0           jdk-17.0.6+10_openj9-0.36.0   about 1 year ago
jdk-17.0.5+8_openj9-0.35.0            jdk-17.0.5+8_openj9-0.35.0    about 1 year ago
jdk-17.0.4.1+1_openj9-0.33.1          jdk-17.0.4.1+1_openj9-0.33.1  about 1 year ago
jdk-17.0.4+8_openj9-0.33.0            jdk-17.0.4+8_openj9-0.33.0    about 1 year ago
jdk-17.0.3+7_openj9-0.32.0            jdk-17.0.3+7_openj9-0.32.0    about 1 year ago
jdk-17.0.2+8_openj9-0.30.0            jdk-17.0.2+8_openj9-0.30.0    about 2 years ago
jdk-17.0.1+12_openj9-0.29.1           jdk-17.0.1+12_openj9-0.29.1   about 2 years ago

I wonder if there is a way of naming the releases to avoid this behaviour? I haven't seen it happening on other repos

[pican79]

I also noticed this weird behavior.

What's interesting is the semeru17-certified-binaries repo doesn't have this issue, even though the releases have similar names:
https://github.com/ibmruntimes/semeru17-certified-binaries/releases

[AdamBrousseau]

Seems to be a GitHub problem.
https://github.com/orgs/community/discussions/8226
I think this latest 17.0.10 OE release is the only one out of order. As @pican79 mentioned, the CE repo is correct, as is the order of the other OE/CE jdk version repos. I am curious what it will look like when we release 17.0.11 in a few weeks. I have tried a few workarounds mentioned in that discussion but can't seem to fix it.

[AdamBrousseau]

Not surprisingly, 17.0.11 is ahead of 17.0.10 but after the other 17 versions.