On Wed, Nov 30, 2016 at 1:24 PM, Karl Tomlinson <[email protected]> wrote:
> Gregory Szorc writes: > > > On Wed, Nov 30, 2016 at 12:40 PM, Karl Tomlinson <[email protected]> > wrote: > >> When history is rewritten, is there a way to view the original > >> history through the web interface, so that autoland tinderbox > >> builds can be used to find regression ranges? > > > > No. Rewritten changesets should result in "replacement" automation > results. > > So there should be no loss of automation coverage. So there should be no > > major impact to things like regression range searching. > > Are the "replacement" results tests re-run with builds from the > new revisions? > Yes. > > Or are the results copied from results for the previous revisions? > No. > If so, is this detectable in treeherder somehow? > Making treeherder react to rewrites is one of the blockers for bug 1288845. > > >> https://hg.mozilla.org/integration/autoland/rev/04a37d89a469 > >> says "An error occurred while processing your request" > > > > This is likely bug 1321344. Ugh, it looks like that issue is more > > widespread than I thought. I'll try to fix it in the next hour. > > I thought the "no" answer above would indicate changesets that > have been rewritten would no longer be visible through the URLs > such as above, but would you expect rewritten changsets to remain > visible on these URLs (if bug 1321344 were fixed)? > Rewritten changesets will HTTP 404. (Although this behavior may change in future versions of Mercurial.) > > >> Is it necessary to pull autoland and use the evolve extension? > > > > It isn't necessary to have evolve enabled to pull the autoland repo. But > if > > evolve is enabled, rewritten changesets that you have pulled from the > > autoland repo will disappear automatically. Otherwise your local repo > will > > keep accumulating old heads now have since been rewritten on the autoland > > repo. > > Thanks. Just to be clear, will pulling from autoland pull old > heads that have been rewritten? > No. The server does not expose rewritten changesets. Its similar to how Git doesn't expose commits that have no refs encasing them: the data is still there but since it is obsolete it isn't accessible via normal means. > > (i.e. I'm not clear whether the accumulating old heads need to be > pulled before the rewrite.) > If you pull a changeset that is later rewritten, you will accumulate heads (the heads rewrites occurred on) unless evolve hides them automagically. _______________________________________________ dev-platform mailing list [email protected] https://lists.mozilla.org/listinfo/dev-platform
