| Age | Commit message (Collapse) | Author |
|---|
|
For some reason just now we were getting a successful build
without a completed time. That's fixed itself now but I might as
well add the code I wrote to handle it anyway.
Probably a glitch on Buildbot's side.
Change-Id: I049da16cf126865ce7d054eca8677da10a3377dc
|
|
libcxx is no longer doing post-commit builds of the main branch.
Pre-commit builds continue but there's no need for us to monitor them,
folks will ping us if there are issues.
Change-Id: I063ec2a22b2c200da5d047bf80998ba4c884df68
|
|
This reverts commit 7f064d26d5fb6c63b7a228cc3416aada5dc919c3.
Tests are now passing again.
Change-Id: I7f4ab51f1a7400e91a0588eeafa76f42ce790f1d
|
|
While https://github.com/llvm/llvm-project/issues/68987 is fixed.
Change-Id: Ic127a7820a91c93796734fff6c3d8da2b22ac4ed
|
|
Change-Id: Ia1dc4047c22f2d8e5a57f0b64e4366e0f114c42d
|
|
We used to ask for all builds and filter from there, but now
that pre-commit llvm PRs are doing many more builds, we aren't
finding a finished main build in the 100 entries we get back
by default.
Instead of filtering afterwards, change the initial query to be
main only.
It's still possible there would be 100 builds without a finished
one, but if that ever happens we can make use of the pagination
feature to get more until we find one.
Change-Id: Ia9e5d090ad86c8257b14be0a2dd81f1237b9233a
|
|
This is what is replacing the latest clang bot.
Change-Id: I5d8baa84f045cd44a862c3b7471fc03d569f7c3e
|
|
This will return shortly as flang-aarch64-libcxx.
Change-Id: I77fd0b90df7491b3a4a6a69f9e9dfa7ec1c31fc7
|
|
Change-Id: Ib188195ee30af70eeff05d9b988e611a914f3ba1
|
|
Change-Id: I35727cfb436098fb05a5d991a719cf5cd0301413
|
|
So originally I was going to cache all the API responses as JSON,
which would be great if we want to test this thing more later, however:
1. Buildkite uses a wrapper library, so it'd take some hacking.
2. JSON can't encode anything but strings as keys.
So instead I've cached the status data using Python's pickle
just after we've read from both sources. This means you can test
changes to the HTML generation in seconds.
Disadvantge being that the pickle object can't be hand constructed.
Switching to JSON or YAML would mean we have to change a bunch of our
keys though, so I don't see the point in that at the moment.
There is one new optional option, --cachefile. If the path exists,
load from it, if it doesn't read the status and save to it.
Change-Id: I45e6edc875271bf258f66655e03d933cbbf02dee
|
|
If we're going to put "Yes" in the column, save the step
of going to the builder and clicking on the latest build.
Change-Id: I0e5a0024a2d73279a0590f4a290aaea73cd7d943
|
|
This is being promoted to normal, and with it goes our last silent
bot (for now).
Change-Id: I8f95319d6fec232f112122cfed247384e60d86ef
|
|
And the old Linaro copy of it that was hanging around.
https://linaro.atlassian.net/browse/LLVM-715 is the plan to replace
this eventually.
Change-Id: Ic0e014ca9f00e2458a0bc2e1196dba87f2f68e16
|
|
This is running on a small Graviton 3 instance. Results should
match the normal bot bar a few failures I have fixes in review for.
Change-Id: Ib9d1855f73e3ff765a375f42c919e81c25cc0a8e
|
|
These have been failing for a long time and aren't providing any value.
https: //linaro.atlassian.net/browse/LLVM-715 will address armv7 in general.
Change-Id: Ia0d74f35979cbbef028d2f1974572647134b60b7
|
|
- The "First Failing" column contains a link to the first failed
build, if the bot has been failing for several builds.
- The "Failing Since" column shows the time elapsed since that first
failed build finished. This can be useful for detecting bots which
fail for a long time.
This column will be colored in red if that time is >24h.
Change-Id: I19878c922c65f8a9f5548ab8e3ea942a57c5a1f6
|
|
Change-Id: I34d1342d43f32a7104b5d171b02cea32a20889d2
|
|
This is a bot status page, it's pretty obvious from context.
Change-Id: Iff2ab4045ceff325a6827492282c6a6534bfe0ac
|
|
Change-Id: I6997699202bf6a3e4a9ab74bbe6be70a6c2a2dca
|
|
make_table.py was already formatted correctly.
Change-Id: I1917983f9763394d5b370c9c76246572a24bcd3c
|
|
We could import a library to do this but it doesn't seem
worth it just for tables.
This uses a content manager so you don't have to write the
begin and end of each cell/row/table. Each method returns
self so you can just keep calling methods to build what you
need.
Change-Id: I8eef664f41039a9fbc662a148f292efd61866694
|
|
These bots will be running on the silent g3-02/g3-03 workers
that are on the newly remade tcwg-g3-04.
Results should be the same as the normal bots and we can
see how the hardware change effects the utilisation.
Change-Id: Ic414b7ed87a7f09521463f4214037debdfc22d21
|
|
Since we are all over the world and we don't want to have
to remember where the server is.
Change-Id: I72ee99735004acde8ac378c339bc19ae71eb2833
|
|
Previously we printed this for all the sections but the time
is the time we generated the HTML not the time we polled the API,
so it didn't need to be per category.
It's at the top so you can tell at a glance if there's an issue
with the script that means it can't update the page.
It's a bit awkward to place it because it needs to go in a table
with content so that it aligns nicely. This leads to a "bug" where
if you had nothing to print, you'd see no "Generated" line, but I
am ok with that.
If the status monitor was blank, I'd have bigger problems than
what time it was made.
Change-Id: Idd9257f75a4c4cbf1be66c3c758f464b47dd3e35
|
|
I've not once cared what the actual number of the build was.
If you want to see where it sits relative to the others, click
the link instead.
Change-Id: I5196a9cee3e782cfe726bf8f5c2ed1c243d1687e
|
|
This column tells you if there is a newer build than the reported
last finished build. This is useful when that reported build has
been cancelled or was a long time ago.
If there's a build in progress, fine, if not, we may need to restart
the agent.
I'm assuming that 'complete' will be True even for a build
that was cancelled or otherwise failed. The docs aren't super
clear on that.
https://docs.buildbot.net/latest/developer/raml/build.html
This column is not coloured red or green because it's up to
the bot maintainer to decide if there being a running build
is required.
Flang hasn't build in a few hours? Fine, it's a low traffic
project. armv8-quick had a cancelled build 2 hours ago and has
no running build? Time to investigate.
Change-Id: I00a49772febe11659e1687b097daf6c9f9651012
|
|
Change-Id: I96dd7c6b2e1e73f706818f4abc3e749095a33109
|
|
These are flang but not compiler-rt bots, so they go in a few categories.
Change-Id: I795c2d6a790b50b73214b9d3ce404499bd3238ad
|
|
These are now running on the AWS Graviton 3 instances.
Change-Id: I6bbf142280d8439337231da97523ace7bcf1233e
|
|
This will move to silent with the other 3 sve bots,
before going to normal later.
Change-Id: I06f1353363d74aae0a88d18f419bc6788b571bec
|
|
I have not once used this column, it is far easier to click through
to the build and look at the changes there.
Change-Id: Icd45199aad86446a438350d7316eb508c58ac965
|
|
This reverts commits:
d32aa9bfdfe84a61a2a9f97c3edfbbf41dc14eb5
004bb870f3a5cf1c8e2184462ede3fb946fe9b0c
a72e680a3e631038d4a7c7f2d9e52bb76d7b323b
b44979c9b2189a3d9ee870af690e20d6a7191548
aarch64-full-2stage, armv8-lld-stage and msvc-2stage
were all testing the compiler-rt runtimes build.
This worked and will be applied on the next normal
master restart.
Change-Id: I0be442b994202c110b2f498f97db11e51bcdeef6
|
|
For more testing of compiler-rt in the runtimes build.
Change-Id: I59828ae561f13412db0e25f88cf62fc98ddc2948
|
|
While I test the compiler-rt runtimes build.
Change-Id: I0cdd48e84a99961b8254de067dc2ba460b5bf2b6
|
|
It's still a flang bot, but on silent temporarily.
Change-Id: I219eeff5f4f3e6dd77016c3af86bda6f98a90931
|
|
This bot is testing building compiler-rt in RUNTIMES,
will move back to normal if that works out.
Change-Id: If019ed5380f3cac73ab29587c58077c8510add68
|
|
This reverts commit c42cbacb39bc9ec897c2b8026c6ace1090e74856.
The breaking commit has been reverted.
Change-Id: I8276dfae47668a4d0aab7796b3cc9c6e599f5baf
|
|
Change-Id: Ifc613073d70da6b83f79fbb8d76c1ff987ee5509
|
|
The host compiler is crashing on something.
Change-Id: I79e73d8cc01403eeef341a6127bc845347414fba
|
|
clangd tests have been flaky in the past week.
Change-Id: I698523d04374d563424b95355dbbecd26669701d
|
|
Change-Id: I4653770bfb187f9d5f91921f2018bc0b5f9df6cd
|
|
These will come back as normal bots at some point in the future.
Change-Id: I490d38d151e535a47a2cc0aa3f481e69c86abdf9
|
|
This will allow us to share this code without
a bunch of one off agreements.
I've chosen MIT because:
* We don't intend to sell it.
* We don't require changes to be contributed back.
(and don't have a great way to accept them anyway)
* We don't want to make any claims about its functionality
being perfect (no warranty).
* Companies see MIT as a friendly license and that's
going to be most of the small audience for this.
* This code does not have any patentable material
that would point us towards some LLVM style Apache license.
Change-Id: I0c6afa0417af9cb281e50cb353ac48d9f3a1900e
|
|
Just because I was snooping around in the Chrome
console and saw a warning that there was none.
This prevents some non-standad rendering apparently,
though I'm sure we don't use enough things to notice.
Change-Id: Ifc43b72d6cab1104689b31704965896af1ae2964
|
|
The .ico files are the only thing in this folder I can't trace
the origin of.
So to add a license file I need to remove them or find new ones.
* The current ones don't work in Chrome (need to be at least 48x48).
* I don't personally use the favicon even if it's there.
* You'd have to have an auto-reload feature to have it update
without you going to the page yourself.
* If it is red, you're going to want to glance at what is
red.
Either way you read the page and the favicon isn't adding anything.
So I'm removing the whole feature. It can come back if
someone wants to find some properly licensed icons.
Change-Id: I220dd7b0c40e46aa37e912145dbd77dc7cb26a68
|
|
This removes a lot of now incorrect information and dead links.
The description of the JSON format has been updated also.
Change-Id: I9001baa2e9af5f64ea031e8b4dacf64337f72e8c
|
|
arvm8-quick replaces armv7-quick.
Flang bot was stable over the weekend after fixes were made.
Change-Id: I597d08d4004c7ba32bcafe51e27f1b7a96086f56
|
|
While some potential race conditions are sorted out.
Change-Id: Id3efaa53cd303fd679b50552e946be9937ee8c94
|
|
I am renaming clang-armv7-quick to armv8 since
that's what it actually builds.
Until the build master restarts we'll be running
a worker for both. Once the switch happens,
I'll remove the armv7 instances.
Change-Id: If4f70e5d18aeb359487d986ab3cac59b43b41770
|
