Bazel: Release 0.5

[hermione521]

Quote from last release manager: hopefully this time it can be 0.5!

[meteorcloudy]

In https://bazel-review.googlesource.com/#/c/9461/, bazel_bootstrap_distfile_test is disabled temporarily on Windows due to https://github.com/bazelbuild/bazel/issues/2708

We need to test it manually before release 0.5

[hlopko]

Thanks for letting me know!

[meteorcloudy]

Please cherry-pick https://github.com/bazelbuild/bazel/pull/2862 if it's not in 0.5. Our users depend on it.

[buchgr]

How is it decided / Who decides what gets into 0.5?

Having bcd23553f38f54fd4846aa507c827a4ee40cfab4 in 0.5 would also be nice, as this is a blocker for google/protobuf.

[hlopko]

It depends on when we will do the cut. We still wait for #1666, then we'll wait for a green ci, then we will do the cut. So https://github.com/bazelbuild/bazel/commit/bcd23553f38f54fd4846aa507c827a4ee40cfab4 will definitely be in the 0.5.

[buchgr]

Makes sense. thanks @mhlopko :)

[damienmg]

FTR 1fb094ecd573b77159f3bc73553ce0750152d2d6 needs to be in the release (regression detected inside google)

[damienmg]

I suggest cutting the release at f3ae88ee043846e7acdffd645137075a4e72c573 since there is only a handful of breakage that needs investigation and no clear regression so far. We can cherry-pick change needed to stabilize the release.

[damienmg]

Note that the release testing is not very good so it would be good to run the docker test on a workstation and send the release branch as a gerrit code review to get bazel test coverage for it.

[buchgr]

@laszlocsomor just informed me that he found 4 very serious bugs in his protobuf patch [1]. We included this patch when updating protobuf in 2b49f678a7dfaf27afc0e44ed7a65b00ec06d413. So we should not include the updated protobuf in the new release (or at least not without updating to laszlo's latest patch).

It looks like the protobuf updated happened before f3ae88e but after 1fb094e.

[1] https://github.com/google/protobuf/pull/2969

[rahul-malik]

@damienmg - Is 0.5 going to include https://github.com/bazelbuild/bazel/issues/1666?

This would be very helpful for distribution of Bazel updates to our team.

[damienmg]

Yes it will include it, the bug isnt closed yet because we still need to
test the release part of it.

Jakob: we need to rollback all those change then :/

On Wed, May 3, 2017, 8:20 PM Rahul Malik [e-mail] wrote:

@damienmg https://github.com/damienmg - Is 0.5 going to include #1666
https://github.com/bazelbuild/bazel/issues/1666?

This would be very helpful for distribution of Bazel updates to our team.

—
You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub
https://github.com/bazelbuild/bazel/issues/2692#issuecomment-298993706,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADjHf8krLlqwjTkUZNhhGm7hpw6wIy1sks5r2MVsgaJpZM4MfgP4
.

[buchgr]

@damienmg :-( ... let's talk with laszo tomorrow before the release.

[laszlocsomor]

4 very serious bugs

The situation isn't so severe. See the commit message for full context: https://github.com/google/protobuf/pull/2969/commits/c4c8806e288915f903f5c2b09a2205b740ffd0d2

[abergmeier-dsfishlabs]

The situation isn't so severe.

That might be true.
Be aware that due to increasingly fewer Bazel releases/months people might stay on that release for quite some time.
What I am asking is probably - will there be 0.5.x before 0.6?

[damienmg]

Yes 0.5.1 will come soon, 0.5.0 cycle was just too slow.

On Thu, May 4, 2017, 10:45 AM Andreas Bergmeier [e-mail]
wrote:

The situation isn't so severe.

That might be true.
Be aware that due to increasingly fewer Bazel releases/months people might
stay on that release for quite some time.
What I am asking is probably - will there be 0.5.x before 0.6?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/bazelbuild/bazel/issues/2692#issuecomment-299128474,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADjHf4bfkLSFAUso3Gr_xqBuoWTPgyTMks5r2ZBDgaJpZM4MfgP4
.

[rahul-malik]

Is there a plan to have a release candidate available before officially shipping 5.0 to incorporate feedback / fix issues before the final release?

[damienmg]

we always do release candidate. We bake them for at least a week.

On Thu, May 4, 2017 at 4:10 PM Rahul Malik [e-mail] wrote:

Is there a plan to have a release candidate available before officially
shipping 5.0 to incorporate feedback / fix issues before the final release?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/bazelbuild/bazel/issues/2692#issuecomment-299196383,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADjHf3DKojAxNzhn7vfIcKRn1ixE9VU4ks5r2dxngaJpZM4MfgP4
.

[hlopko]

Creating RC3, for the info:
base: f3ae88ee043846e7acdffd645137075a4e72c573
cherrypicks:
c58ba098526b748f9c73e6229cafd74748205aa1
0acead4ea3631240659836ce6ecd6d7f67fd352b
d0242ce4a87929f2528f4602d0fb09d1ccfcea94
d953ca8b87a46decbce385cebb446ae0dd390881

[petemounce]

@mhlopko I had a look for https://storage.googleapis.com/bazel/0.5.0/rc3/index.html to publish a chocolatey package for the RC, but I get a NoSuchKey from it. Is that still the correct URL? I also didn't see a note in bazel-dev...?

[damienmg]

It is not published yet. The CI system has to build the artifact and
publish them. A mail is send to the mailing list right after it is done.

On Fri, May 5, 2017 at 11:46 AM Peter Mounce [e-mail]
wrote:

@mhlopko https://github.com/mhlopko I had a look for
https://storage.googleapis.com/bazel/0.5.0/rc3/index.html to publish a
chocolatey package for the RC, but I get a NoSuchKey from it. Is that
still the correct URL? I also didn't see a note in bazel-dev...?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/bazelbuild/bazel/issues/2692#issuecomment-299424506,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADjHfydMc8audsccoI_6ckwCmahI22rcks5r2u_ugaJpZM4MfgP4
.

[hlopko]

Wow @petemounce you''re quick! I just created it minutes ago, it should show up soon.

[petemounce]

If you point me at the release automation, perhaps I can build the chocolatey package publishing into it? Assuming there's a Windows box that can run powershell to build & test-install the package, that is?

[damienmg]

Hi peter, this is a bit tricky. How are you building the chocolatey package?

[petemounce]

Via scripts/packages/chocolatey/build|test.ps1. choco itself apparently runs on linux, but no binaries for that are distributed according to a stackoverflow answer.

I'd be fine to rewrite the build/test/release scripts into golang, actually, if that makes this easier - I'd use them internally too.

[damienmg]

We can run steps on windows that's not an issue. I'll try to add it as part of the revamp of our ci.

[Helcaraxan]

@damienmg what is happening to the RC3 image? From what I can see the page given above by @petemounce (https://storage.googleapis.com/bazel/0.5.0/rc3/index.html) still returns:

<Error>
  <Code>NoSuchKey</Code>
  <Message>The specified key does not exist.</Message>
</Error>

[damienmg]

https://storage.googleapis.com/bazel/0.5/rc3/index.html the release number
was set to 0.5 not 0.5.0, we will correct it for the next release candidate.

On Mon, May 8, 2017 at 12:39 PM Duco van Amstel [e-mail]
wrote:

@damienmg https://github.com/damienmg what is happening to the RC3
image? From what I can see the page given above by @petemounce
https://github.com/petemounce (
https://storage.googleapis.com/bazel/0.5.0/rc3/index.html) still returns:

NoSuchKey
The specified key does not exist.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/bazelbuild/bazel/issues/2692#issuecomment-299833187,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADjHf76Khj3pWVrKPNL85nzOEoi5EukDks5r3vDcgaJpZM4MfgP4
.

[petemounce]

I've published 0.5.0-rc3 by local hacks to the scripts. Nuget only accepts semver-compliant version numbers, so wouldn't accept 0.5-rc3. Maybe some validation to the release-creation script - if you point me at that I could add it?

https://chocolatey.org/packages/bazel/0.5.0-rc3

Diff

λ  git d
diff --git i/scripts/packages/chocolatey/build.ps1 w/scripts/packages/chocolatey/build.ps1
index 3d479ad65..263de241e 100644
--- i/scripts/packages/chocolatey/build.ps1
+++ w/scripts/packages/chocolatey/build.ps1
@@ -13,7 +13,7 @@ if ($mode -eq "release") {
   $tvUri = "https://github.com/bazelbuild/bazel/releases/download/$($tvVersion)/$($tvFilename)"
   $tvReleaseNotesUri = "https://github.com/bazelbuild/bazel/releases/tag/$tvVersion"
 } elseif ($mode -eq "rc") {
-  $tvVersion = "$($version)-rc$($rc)"
+  $tvVersion = "$($version).0-rc$($rc)"
   $tvFilename = "bazel-$($version)rc$($rc)-windows-x86_64.zip"
   $tvUri = "https://storage.googleapis.com/bazel/$($version)/rc$($rc)/$($tvFilename)"
   $tvReleaseNotesUri = "https://storage.googleapis.com/bazel/$($version)/rc$($rc)/index.html"
diff --git i/scripts/packages/chocolatey/test.ps1 w/scripts/packages/chocolatey/test.ps1
index a2349351c..39dc023dd 100644
--- i/scripts/packages/chocolatey/test.ps1
+++ w/scripts/packages/chocolatey/test.ps1
@@ -6,9 +6,9 @@ param(

 choco uninstall bazel --force -y
 if ($prerelease) {
-  choco install ".\bazel.$($version).nupkg" --verbose --debug --prerelease --force -y -s $sources
+  choco install ".\bazel.$($version).0-rc3.nupkg" --verbose --debug --prerelease --force -y -s $sources
 } else {
-  choco install ".\bazel.$($version).nupkg" --verbose --debug --force -y -s $sources
+  choco install ".\bazel.$($version).0-rc3.nupkg" --verbose --debug --force -y -s $sources
 }

 if ($LASTEXITCODE -ne 0)

[hlopko]

That would be very helpful! This is the script: https://github.com/bazelbuild/bazel/blob/master/scripts/release/release.sh

[hlopko]

I created rc4, now in a correct branch named 0.5.0:
https://storage.googleapis.com/bazel/0.5.0/rc4/index.html

base: f3ae88e
cherrypicks:
c58ba09
0acead4
d0242ce
d953ca8
755669fb5de1f4e762f27c19776cac9f410fcb94
6f041661ca159903691fcb443d86dc7b6454253d
720561113bfa702acfc2ca24ce3cc3fd7ee9c115
511c35b46cead500d4e76706e0a709e50995ceba

And we already know we will have at least one more rc.

[damienmg]

Some updates:
There 2 release blocker open right now, @philwo pinged me about another bug that surfaced inside of Google (he will update the thread with more details later) and we are still waiting for a green from some of users of Bazel (including TensorFlow).

The good part is rc6 is bootstrapping fine in homebrew :p

[petemounce]

I've pushed RCs that I know about:

• https://chocolatey.org/packages/bazel/0.5.0-rc4
• https://chocolatey.org/packages/bazel/0.5.0-rc5
• https://chocolatey.org/packages/bazel/0.5.0-rc6

[petemounce]

@damienmg @laszlocsomor et al - what, if anything, would you like me to do to the chocolatey package to represent the fact that there are now 3 flavours of bazel possible to get as I can see:

• current one
• one that doesn't have jdk7 bundled
• one that is msysless
  ?

I could add a flag to the installer to specify which zip to download, and therefore which checksum to use, but that would be a bit of work. Fine to do it, but thought I'd check first. At work, we'll be trying out the msysless one, I think, so will want to make that easy to install.

What is the lifetime of the different flavours expected to be? Is this kind of scenario one that would be beneficial to support now and/or in future?

Caveat: the package should only end up providing one bazel.exe - any other exes should be named for the flavour, so that they can coexist on a system and not have the user juggle the PATH to choose which one to use, I think.

[laszlocsomor]

@petemounce : I replied on https://github.com/bazelbuild/bazel/issues/3006
Also, thank you for pushing the RCs!

[dslomov]

Please cherrypick (infrastructure changes):
https://github.com/bazelbuild/bazel/commit/eb565f408e03125e92d42b00756e519795be6593
https://github.com/bazelbuild/bazel/commit/39f328cf392056618d1a3ead4835a138b189a06d
https://github.com/bazelbuild/bazel/commit/624802893f4fe72118f00a78452605d41a2e1c6f

[philwo]

@mhlopko Can we put a big note in the release notes regarding sandboxing:

"Bazel's Linux sandbox no longer mounts an empty tmpfs on /tmp, instead the existing /tmp is mounted read-write. If you prefer to have a tmpfs on /tmp for sandboxed actions for increased hermeticity, please use the flag --sandbox_tmpfs_path=/tmp."

[hlopko]

Please celebrate another rc :) Including big note and 3b08f774e7938928e3a240a47a0a7554cdc8d50b :)

[meteorcloudy]

Thank you, Marcel!! \o/

[ittaiz]

@iirina will https://github.com/bazelbuild/bazel/commit/7e91ad631a660ae12a2b8dcfb8ba42574a7a76d3 get in 0.5.0?

[iirina]

It wasn't planned for, but if you have a strong opinion about it being included you can ask @mhlopko if he would create another RC. But since this is already the 8th RC and a release should be done soon, I'm not very sure about it.

[philwo]

If you do another round of cherry-picks, please pick this in: https://github.com/bazelbuild/bazel/commit/db5e06a29fccd31ad8ae13e7d271509807d87d7c

I promised it to a user here: https://stackoverflow.com/questions/43849651/how-to-lock-down-the-bazel-filesystem-sandbox/43897863#43897863

But, to be honest, I think we should call it a day and release this if it's stable and then quickly do a 0.5.1 instead of dragging this out even longer...

[ittaiz]

@philwo @iirina realistically when can we expect 0.5.1?
If 0.5.1 can be cut in a few days (week or two) then of course no objection.
If it's a month from now then I'd say pretty please :)

[damienmg]

We need another release out by the end of June and that's a hard deadline
for us so we will cut 0.5.1 a few days after 0.5.0 is released.

+1 for Philip's words

On Tue, May 23, 2017, 6:56 AM Ittai Zeidman [e-mail]
wrote:

@philwo https://github.com/philwo @iirina https://github.com/iirina
realistically when can we expect 0.5.1?
If 0.5.1 can be cut in a few days (week or two) then of course no
objection.
If it's a month from now then I'd say pretty please :)

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/bazelbuild/bazel/issues/2692#issuecomment-303406088,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADjHf6Ao_zuo3dQlc80sKzUfzO7qNkGsks5r8uWpgaJpZM4MfgP4
.

[petemounce]

I tried to push rc7 (https://github.com/bazelbuild/bazel/issues/2692#issuecomment-303335602), but it doesn't appear to have arrived in the bucket yet? That is, https://storage.googleapis.com/bazel/0.5.0/rc7/index.html -> NoSuchKey.

[hlopko]

Oh sorry about that, I messed 7 up and didn't push it, and only realized so when I pushed 8... So there's no RC 7...

[petemounce]

Ah, no worries. Thanks.

https://chocolatey.org/packages/bazel/0.5.0-rc8 has been published.

[philwo]

@mhlopko A release critical bug was found in our internal version in which the linux-sandbox is completely non functional (can't even execute /bin/true), cf. b/62022773. The bug does not seem to affect Bazel 0.5.0, but please let me verify this manually before finally releasing Bazel 0.5.0. :( Sorry for the mess...

Amazingly, the bug is present in Bazel's HEAD and causes Bazel to automatically fall back to non-sandboxed execution, but nothing in our CI is catching it. It's totally green and saying that sandboxed tests are passing - WTF.

[damienmg]

IIUC from the internal bug, the causing CL is not in the release branch. All the test from Google teams have come back green, no regression reported by external users, seems like we are good to push 0.5.0rc9 as the 0.5.0 release tomorrow?

[petemounce]

https://chocolatey.org/packages/bazel/0.5.0-rc9 has been published.

[philwo]

Damien: OK, I think we can then go ahead with the release.

[damienmg]

This was released but there is a regression: https://github.com/bazelbuild/bazel/issues/3063 :(

Anyway closing this bug.

[petemounce]

https://chocolatey.org/packages/bazel/0.5.0 is published, awaiting automated review.