|
|
Created:
5 years, 6 months ago by Xianzhu Modified:
5 years, 6 months ago CC:
blink-reviews Base URL:
svn://svn.chromium.org/blink/trunk Target Ref:
refs/heads/master Project:
blink Visibility:
Public. |
DescriptionIgnore builders without result
The newly added "WebKit Mac10.10" builder has no result yet, and caused
crash of rebaseline script.
BUG=502141
TBR=joelo@chromium.org,dpranke@chromium.org
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=197432
Patch Set 1 #
Messages
Total messages: 11 (0 generated)
Message was sent while issue was closed.
Committed patchset #1 (id:1) manually as 197432 (presubmit successful).
Message was sent while issue was closed.
what happens if we fail to get the results from a master we need to do the rebaselining correctly?
Message was sent while issue was closed.
On 2015/06/20 00:16:50, Dirk Pranke wrote: > what happens if we fail to get the results from a master we need to do the > rebaselining correctly? NetworkTransaction.run() will retry the request until a correct result or 404 (will return None) is got, or timeout (will raise exception).
Message was sent while issue was closed.
On 2015/06/20 00:42:28, Xianzhu wrote: > On 2015/06/20 00:16:50, Dirk Pranke wrote: > > what happens if we fail to get the results from a master we need to do the > > rebaselining correctly? > > NetworkTransaction.run() will retry the request until a correct result or 404 > (will return None) is got, or timeout (will raise exception). That's not quite what I mean. What I mean is, if we really need to get a result from, say, Mavericks in order to make sure the graph is updated correctly, if we return None here and bypass that builder, does that mean that we'll delete the [ NeedsRebaseline ] entries and rebaseline incorrectly?
Message was sent while issue was closed.
On 2015/06/20 00:59:48, Dirk Pranke wrote: > On 2015/06/20 00:42:28, Xianzhu wrote: > > On 2015/06/20 00:16:50, Dirk Pranke wrote: > > > what happens if we fail to get the results from a master we need to do the > > > rebaselining correctly? > > > > NetworkTransaction.run() will retry the request until a correct result or 404 > > (will return None) is got, or timeout (will raise exception). > > That's not quite what I mean. > > What I mean is, if we really need to get a result from, say, Mavericks in order > to make sure the graph > is updated correctly, if we return None here and bypass that builder, does that > mean that we'll delete > the [ NeedsRebaseline ] entries and rebaseline incorrectly? I'm confused: why would someone rebaseline for a builder (e.g. Mac10.10) which doesn't exist or has not generate any result? Another question is: when will Mac10.10 generate the first result? The CLs were TBR'ed because I didn't want rebaseline-o-matic to be blocked too long. I can revert them when we have Mac10.10 setup.
Message was sent while issue was closed.
On 2015/06/20 02:57:31, Xianzhu wrote: > On 2015/06/20 00:59:48, Dirk Pranke wrote: > > On 2015/06/20 00:42:28, Xianzhu wrote: > > > On 2015/06/20 00:16:50, Dirk Pranke wrote: > > > > what happens if we fail to get the results from a master we need to do the > > > > rebaselining correctly? > > > > > > NetworkTransaction.run() will retry the request until a correct result or > 404 > > > (will return None) is got, or timeout (will raise exception). > > > > That's not quite what I mean. > > > > What I mean is, if we really need to get a result from, say, Mavericks in > order > > to make sure the graph > > is updated correctly, if we return None here and bypass that builder, does > that > > mean that we'll delete > > the [ NeedsRebaseline ] entries and rebaseline incorrectly? > > I'm confused: why would someone rebaseline for a builder (e.g. Mac10.10) which > doesn't exist or has not generate any result? > Another question is: when will Mac10.10 generate the first result? > > The CLs were TBR'ed because I didn't want rebaseline-o-matic to be blocked too > long. I can revert them when we have Mac10.10 setup. Ah, sorry for breaking things! I'll get the builder up asap. It's not clear to me if this will cause auto-rebaseline to rebaseline incorrectly if a builder for a current version is down. I'll run a test when I get the chance. We should probably handle this properly so future versions don't break the script when added.
Message was sent while issue was closed.
On 2015/06/22 10:03:54, joelo wrote: > On 2015/06/20 02:57:31, Xianzhu wrote: > > On 2015/06/20 00:59:48, Dirk Pranke wrote: > > > On 2015/06/20 00:42:28, Xianzhu wrote: > > > > On 2015/06/20 00:16:50, Dirk Pranke wrote: > > > > > what happens if we fail to get the results from a master we need to do > the > > > > > rebaselining correctly? > > > > > > > > NetworkTransaction.run() will retry the request until a correct result or > > 404 > > > > (will return None) is got, or timeout (will raise exception). > > > > > > That's not quite what I mean. > > > > > > What I mean is, if we really need to get a result from, say, Mavericks in > > order > > > to make sure the graph > > > is updated correctly, if we return None here and bypass that builder, does > > that > > > mean that we'll delete > > > the [ NeedsRebaseline ] entries and rebaseline incorrectly? > > > > I'm confused: why would someone rebaseline for a builder (e.g. Mac10.10) which > > doesn't exist or has not generate any result? My concern is the same one Joel metnions, below: what happens if a bot that is normally functioning fine starts returning 404s? I'm less concerned about bots that have never worked (like 10.10 currently).
Message was sent while issue was closed.
On 2015/06/22 22:59:49, Dirk Pranke wrote: > On 2015/06/22 10:03:54, joelo wrote: > > On 2015/06/20 02:57:31, Xianzhu wrote: > > > On 2015/06/20 00:59:48, Dirk Pranke wrote: > > > > On 2015/06/20 00:42:28, Xianzhu wrote: > > > > > On 2015/06/20 00:16:50, Dirk Pranke wrote: > > > > > > what happens if we fail to get the results from a master we need to do > > the > > > > > > rebaselining correctly? > > > > > > > > > > NetworkTransaction.run() will retry the request until a correct result > or > > > 404 > > > > > (will return None) is got, or timeout (will raise exception). > > > > > > > > That's not quite what I mean. > > > > > > > > What I mean is, if we really need to get a result from, say, Mavericks in > > > order > > > > to make sure the graph > > > > is updated correctly, if we return None here and bypass that builder, does > > > that > > > > mean that we'll delete > > > > the [ NeedsRebaseline ] entries and rebaseline incorrectly? > > > > > > I'm confused: why would someone rebaseline for a builder (e.g. Mac10.10) > which > > > doesn't exist or has not generate any result? > > My concern is the same one Joel metnions, below: what happens if a bot > that is normally functioning fine starts returning 404s? I'm less concerned > about > bots that have never worked (like 10.10 currently). I don't know much about the servers. With this concern, I'll revert the changes. Can I do it now without breakage?
Message was sent while issue was closed.
On 2015/06/22 23:02:17, Xianzhu wrote: > I don't know much about the servers. With this concern, I'll revert the changes. > Can I do it now without breakage? I wouldn't worry about it for now. This is probably a mostly theoretical concern. Let's just leave things the way they are and get 10.10 up and running and let Joel worry about the long run ;).
Message was sent while issue was closed.
Is it possible to run rebaseline-o-matic with local changes? If yes, I'll make local changes to workaround failures instead of submitting such changes.
Message was sent while issue was closed.
Yup, you can have the script hardcoded locally to ignore 10.10 for now. But as Dirk mentioned I wouldn't worry about it too much. I'm filing to request to bring up the 10.10 bot right now. |