>Yocto has a download cache mechanism, so you should not have to fetch
>more than once. Are you throwing away that cache between builds?

Do you mean the DL_DIR or something else as cache mechanism? We do keep DL_DIR between builds, but we're talking about cases where there was a change in the repository.

>the actual problem is slow gerrit/artifactory instances, not

>stuck fetch tasks

Totally agree, but we're not able to fix the slow gerrit/artifactory in a short term.

>The idea of 'retrying' the fetch

Well, it's not like I'm saying this as a solution, usually we just disable jobs until the slowness is gone. But isn't that what we already do with wget in bitbake?

"wget -t 2 -T 30" from bitbake/lib/bb/fetch2/wget.py, where -t is tries, and -T is timeout

Problem is that we get the slowness for git repositories and git does not support such a thing.

Best regards,
Tomasz Dziendzielski