P1: Critical Description
For some reason, 'gerrit staging-ls' for Qt 4.8 takes quite a long time to run:
$ time ssh -oBatchMode=yes -p 29418 codereview.qt-project.org gerrit staging-ls --branch refs/staging/4.8 --destination refs/heads/4.8 --project qt/qt ; time ssh -oBatchMode=yes -p 29418 codereview.qt-project.org gerrit staging-ls --branch refs/staging/4.8 --destination refs/heads/4.8 --project qt/qt ; time ssh -oBatchMode=yes -p 29418 codereview.qt-project.org gerrit staging-ls --branch refs/staging/4.8 --destination refs/heads/4.8 --project qt/qt cdf50d213ec533062995d5ffa32ab14999f9eb2e 22925,2 Mangle QNSApplication into a namespace when Qt is configured as such 9a89d1c9e1018b459b7e5b99eb4e6fba37f22274 32150,2 Fix leak in QDragManager::drag on mac 6d50ae2f70347a3551b978ace26ae6b484e9f1be 35561,2 Fixing font database memoryleak ssh -oBatchMode=yes -p 29418 codereview.qt-project.org gerrit staging-ls 0.01s user 0.00s system 0% cpu 7:38.29 total cdf50d213ec533062995d5ffa32ab14999f9eb2e 22925,2 Mangle QNSApplication into a namespace when Qt is configured as such 9a89d1c9e1018b459b7e5b99eb4e6fba37f22274 32150,2 Fix leak in QDragManager::drag on mac 6d50ae2f70347a3551b978ace26ae6b484e9f1be 35561,2 Fixing font database memoryleak ssh -oBatchMode=yes -p 29418 codereview.qt-project.org gerrit staging-ls 0.01s user 0.00s system 0% cpu 7:47.45 total cdf50d213ec533062995d5ffa32ab14999f9eb2e 22925,2 Mangle QNSApplication into a namespace when Qt is configured as such 9a89d1c9e1018b459b7e5b99eb4e6fba37f22274 32150,2 Fix leak in QDragManager::drag on mac 6d50ae2f70347a3551b978ace26ae6b484e9f1be 35561,2 Fixing font database memoryleak ssh -oBatchMode=yes -p 29418 codereview.qt-project.org gerrit staging-ls 0.01s user 0.00s system 0% cpu 7:43.59 total
At time of writing, it takes 7-8 minutes to run staging-ls on Qt 4.8. qtbase master, for comparison, takes about 20 seconds.
This causes noticeable delays at the start of each CI round of Qt 4.8, since the system takes so long to check whether any changes are staged.
From CI logs, it looks like staging-new-build is similarly slow, but I haven't tested that separately since it's not a read-only command like staging-ls.