[ https://issues.apache.org/jira/browse/HBASE-20070?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16377524#comment-16377524 ]
Sean Busbey commented on HBASE-20070: ------------------------------------- {quote} I think this might not be quite right (mine was also probably wrong). If I am understanding this correctly, this will pass if the hash is in any branch we know about, not just `master`, but I think we only care if it is in `master`. Can we narrow it down more? I'm also not in love with this indent-less syntax for the if contents. {quote} the git command that counted things for PUSHED only traces back commits starting from some reference (in this case the local HEAD), so it should only count things in the {{asf-site}} branch of the hbase-site repo. Which I think is correct? The {{if}} block in my patch should be indented two spaces. It is when I open things in vim at least. {quote} I think we need to document that the stuff that removes ~/.m2 may not work locally but that's specifically to try to avoid Jenkins server problems so it can probably be ignored locally. In fact maybe there is some cleverer way for us to sandbox ourselves, but I am not a Maven expert. {quote} If {{WORKSPACE}} exists, then the temp directory should work. We could move it out of this script though, and just have folks pass in a path for what they want used for the maven local repository. we could then manage the per-workspace repository via jenkins. better? {quote} I think maybe we need to add a –push=true or similar flag to this script so we can do local tests without it always pushing the new site. {quote} Sure. I've been commenting out the push while I have a WIP. I can add a flag for publishing. > website generation is failing > ----------------------------- > > Key: HBASE-20070 > URL: https://issues.apache.org/jira/browse/HBASE-20070 > Project: HBase > Issue Type: Bug > Components: website > Reporter: Sean Busbey > Assignee: Sean Busbey > Priority: Blocker > Attachments: HBASE-20070-misty.patch, HBASE-20070-misty.patch.1, > HBASE-20070.0.patch, HBASE-20070.1.patch, HBASE-20070.2.patch, > hbase-install-log-a29b3caf4dbc7b8833474ef5da5438f7f6907e00.txt > > > website generation has been failing since Feb 20th > {code} > Checking out files: 100% (68971/68971), done. > Usage: grep [OPTION]... PATTERN [FILE]... > Try 'grep --help' for more information. > PUSHED is 2 > is not yet mentioned in the hbase-site commit log. Assuming we don't have it > yet. 2 > Building HBase > Java HotSpot(TM) 64-Bit Server VM warning: ignoring option MaxPermSize=256m; > support was removed in 8.0 > Java HotSpot(TM) 64-Bit Server VM warning: ignoring option MaxPermSize=256m; > support was removed in 8.0 > Failure: mvn clean site > Build step 'Execute shell' marked build as failure > {code} > The status email says > {code} > Build status: Still Failing > The HBase website has not been updated to incorporate HBase commit > ${CURRENT_HBASE_COMMIT}. > {code} > Looking at the code where that grep happens, it looks like the env variable > CURRENT_HBASE_COMMIT isn't getting set. That comes from some git command. I'm > guessing the version of git changed on the build hosts and upended our > assumptions. > we should fix this to 1) rely on git's porcelain interface, and 2) fail as > soon as that git command fails -- This message was sent by Atlassian JIRA (v7.6.3#76005)