GitHub pull request #2476 of commit 6e487a4fcb6c4f2e4b21982def07b84acbc76e8b, no merge conflicts.sha1origin/pr/2476/mergeghprbActualCommit6e487a4fcb6c4f2e4b21982def07b84acbc76e8bghprbActualCommitAuthorDave MarchevskyghprbActualCommitAuthorEmaildavemarchevsky@gmail.comghprbAuthorRepoGitUrlhttps://github.com/davemarchevsky/bcc.gitghprbTriggerAuthorghprbTriggerAuthorEmailghprbTriggerAuthorLoginyonghong-songghprbTriggerAuthorLoginMention@yonghong-songghprbPullId2476ghprbTargetBranchmasterghprbSourceBranchusdt_partial_initGIT_BRANCHusdt_partial_initghprbPullAuthorEmaildavemarchevsky@gmail.comghprbPullAuthorLogindavemarchevskyghprbPullAuthorLoginMention@davemarchevskyghprbPullDescriptionGitHub pull request #2476 of commit 6e487a4fcb6c4f2e4b21982def07b84acbc76e8b, no merge conflicts.ghprbPullTitleAdd allow_partial_usdt_init param so partial init of USDT vector is possibleghprbPullLinkhttps://github.com/iovisor/bcc/pull/2476ghprbPullLongDescriptionLooking for early feedback here before I write tests / flesh this out more.\r\n\r\nConsider this scenario: I'm hoping to attach USDTs to many short-lived processes, let's say 30. I want to gather data from USDT handlers on these processes for 5 seconds. It's guaranteed that at least one of the processes that was alive when I constructed the vector of USDTs to pass into `BPF::init` will not be alive by the time the 5 seconds are up. I'm OK with this and would still like to gather as much data as possible.\r\n\r\nIf the process dies after `USDT::init()` is successful but before `attach_usdt` is called, we're already OK because we can handle each `StatusTuple` independently outside of `bcc`. Similarly, if process dies between `attach_usdt` and `detach_usdt` calls, we're fine for same reason.\r\n\r\nBut if it dies between construction of `USDT` object and `USDT::init()` being called by `BPF::init()`, we currently consider failure to init any `USDT` critical enough to stop the whole `BPF::init()`. \r\n\r\nThis PR adds a `allow_partial_usdt_init` param to `BPF::init` to let users of `bcc` signal that they're OK with `USDT::init()` failures. `USDT::initialized()` getter is added so `bcc` user can check which `USDT`s succeeded / failed.\r\n\r\nOne downside of this initial pass is that the info in `StatusTuple`s for `USDT::init` is lost. So if we have some `USDT::init` failures when `allow_partial_usdt_init` is `true` it'll be hard to get visibility into them. \r\n\r\n/cc @yonghong-song ghprbCommentBodynullghprbGhRepositoryiovisor/bccghprbCredentialsId6d3daf13-69b8-48b1-9c8f-ec5353264113false<a title="Add allow_partial_usdt_init param so partial init of USDT vector is possible" href="https://github.com/iovisor/bcc/pull/2476">PR #2476</a>: Add allow_partial_usdt_init...#12310994857460664bcc-pr #123123false1231012SUCCESS1573685082067https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/123/git123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=fc25/123/123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=fc26/123/123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=fc27/123/123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=fc28/123/123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=ubuntu1604/123/123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=ubuntu1710/123/123https://buildbot.iovisor.org/jenkins/user/dalehamel/my-views/view/all/job/bcc-pr/label=ubuntu1804/123/