Implement Hermetic sandbox with support for hardlinks #13279

frazze-jobb · 2021-03-30T07:47:20Z

Adds linux-sandbox flag:
--experimental_use_hermetic_linux_sandbox - Configure linux-sandbox
to run in a chroot environment to prevent access to files not
mentioned in the bazel rules unless they can be found via
explicitly whitelisted directories using --sandbox_add_mount_pair
create hardlinks instead of symlinks, and fallback to copying.
In case of writes to input files, the build will be aborted.

frazze-jobb · 2021-03-30T07:47:41Z

Related to Issue Make the sandboxed file system more strict: #7313

google-cla · 2021-03-30T07:47:59Z

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project (if not, look below for help). Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed (or fixed any issues), please reply here with @googlebot I signed it! and we'll verify it.

What to do if you already signed the CLA

Individual signers

It's possible we don't have your GitHub username or you're using a different email address on your commit. Check your existing CLA data and verify that your email is set on your git commits.

Corporate signers

Your company has a Point of Contact who decides which employees are authorized to participate. Ask your POC to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the Google project maintainer to go/cla#troubleshoot (Public version).
The email used to register you as an authorized contributor must be the email used for the Git commit. Check your existing CLA data and verify that your email is set on your git commits.
The email used to register you as an authorized contributor must also be attached to your GitHub account.

ℹ️ Googlers: Go here for more info.

google-cla · 2021-03-30T08:39:15Z

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project (if not, look below for help). Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed (or fixed any issues), please reply here with @googlebot I signed it! and we'll verify it.

What to do if you already signed the CLA

Individual signers

It's possible we don't have your GitHub username or you're using a different email address on your commit. Check your existing CLA data and verify that your email is set on your git commits.

Corporate signers

Your company has a Point of Contact who decides which employees are authorized to participate. Ask your POC to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the Google project maintainer to go/cla#troubleshoot (Public version).
The email used to register you as an authorized contributor must be the email used for the Git commit. Check your existing CLA data and verify that your email is set on your git commits.
The email used to register you as an authorized contributor must also be attached to your GitHub account.

ℹ️ Googlers: Go here for more info.

aiuto · 2021-03-31T03:25:09Z

It is not clear we had agreement on answer to #7313, so this PR might be premature.

frazze-jobb · 2021-03-31T12:51:54Z

This is the results from a benchmark on how long one of our builds takes with respective sandbox.

Sandbox type | Average build-time
Process-wrapper | 268.93
Linux-sandbox | 287.26
Hermetic Linux-sandbox | 271.37

justinhorvitz · 2021-03-31T21:05:33Z

I will let philwo handle this review, since I'm not that familiar with sandboxes.

src/main/tools/linux-sandbox-options.cc

larsrc-google

Not quite done with the C++/Bash parts yet, but here's a couple of notes.

src/main/tools/linux-sandbox-pid1.cc

larsrc-google · 2021-04-09T14:18:08Z

Could you fix the tests?

frazze-jobb · 2021-04-12T18:55:16Z

Thanks for your review comments. Regarding the test I thought I knew what to fix, turns out I was wrong.
Can I get the test environment somehow to try and reproduce the problem?

larsrc-google · 2021-04-13T14:03:03Z

Click the Details link at the failing test, it'll show you the log. But that's the log for the entire shard. Click "Artifacts" to see the output from the failing tests. Once you know what tests are failing, run them locally (possibly with ibazel).

frazze-jobb · 2021-04-20T12:56:16Z

I seem to have solved most of the tests, but how do I prevent "buildkite/bazel-bazel-github-presubmit/darwin-openjdk-8-shard-2" from running my hermetic "linux-sandbox" test. I assume linux-sandbox is not supported on that one, and therefore the tests will be failing

larsrc-google · 2021-04-20T14:20:32Z

Looking at the logs, one test is failing because the error message is different, one is passing, and one is failing for actually not being hermetic. The last one is the interesting one. For that, I would suggest doing a platform check like this.

frazze-jobb · 2021-04-26T07:09:07Z

The tests are passing now, I have dealt with @larsrc-google initial review comments. I would appreciate some more review comments @philwo

larsrc-google

Here are a few fixes that make it easier for me to test further revisions, and a couple of clarification questions. All tests pass, though.

src/main/java/com/google/devtools/build/lib/sandbox/LinuxSandboxedSpawnRunner.java

src/main/tools/linux-sandbox-pid1.cc

src/main/java/com/google/devtools/build/lib/sandbox/HardlinkedSandboxedSpawn.java

larsrc-google · 2021-04-26T08:23:27Z

src/test/shell/bazel/bazel_hermetic_sandboxing_test.sh

+
+# For the test to work we need to bind mount a couple of folders to
+# get access to bash, ls, python etc. Depending on linux distribution
+# these folders may vary. Mount all folders in the root directory '/' 


Nit: Remove empty trailing whitespace. If your editor has an option to remove trailing whitespace, I suggest turning that on.

src/main/java/com/google/devtools/build/lib/sandbox/HardlinkedSandboxedSpawn.java

src/main/java/com/google/devtools/build/lib/sandbox/LinuxSandboxedSpawnRunner.java

src/main/java/com/google/devtools/build/lib/sandbox/AbstractSandboxSpawnRunner.java

src/main/java/com/google/devtools/build/lib/actions/FileContentsProxy.java

src/main/tools/linux-sandbox-options.cc

frazze-jobb · 2021-04-29T13:22:33Z

The test failing

public void testUnreadableFileWithNoFastDigest() throws Exception

expects that setting a new modified time should not affect this assert.

p.setLastModifiedTime(10L);
assertThat(valueForPath(p)).isEqualTo(value);

But I just added mtime to the hash/equal/fingerprint and prettyprintt so this is expected. What should I do about this test?

larsrc-google · 2021-04-29T13:53:40Z

@janakdr is probably the one who would know about the mtime handling.

janakdr

@michajlo wrote that test, he may have a better idea. On its face, I don't see that the test is enforcing an important property.

janakdr · 2021-04-29T16:23:15Z

src/main/java/com/google/devtools/build/lib/actions/FileContentsProxy.java

  /**
   * Visible for serialization / deserialization. Do not use this method, but call {@link #create}
   * instead.
   */
  public FileContentsProxy(long ctime, long nodeId) {
    this.ctime = ctime;
+    this.mtime = ctime;


Cleaned this class up in e4cc0b9: would you mind syncing?

michajlo · 2021-04-29T19:24:21Z

That was a while ago... Skimming history I believe I was showing that mtime was significant to equality (the version I introduced was actually asserting not-equal). At some point it got flipped to assert equal when mtime was removed / deemed unnecessary?

philwo

In general LGTM, added a few comments.

philwo · 2021-06-08T14:53:35Z

src/main/tools/linux-sandbox-options.cc

+          // Make sure that the sandbox_root path has no trailing slash.
+          if (sandbox_root.back() == '/') {


What happens when the path has two trailing slashes, e.g. /dev/shm/mysandbox//? (I'm fine with exiting with an error on invalid input like that, but it looks like this wouldn't catch it at all?)

philwo · 2021-06-08T14:54:56Z

src/main/tools/linux-sandbox-pid1.cc

+  if (handle < 0) {
+    DIE("open");
+  }
+  if (close(handle)) {


Could you add < 0 to the condition?

philwo · 2021-06-08T15:01:22Z

src/main/tools/linux-sandbox-pid1.cc

+    // Create a bind mount to /proc
+    if (CreateTarget("proc", true) < 0) {
+      DIE("CreateTarget");
+    }
+    if (mount("/proc", "proc", NULL, MS_REC | MS_BIND, NULL) < 0) {
+      DIE("mount");
+    }


What's the reason that we no longer have to mount a new proc on top of the old one, because the old one still refers to our parent PID namespace, but can just use a bind mount in this case?

I could have a look, this code is taken from old linux-sandbox implementation, but maybe its not needed.

Thanks! I remember this being tricky, but in the end we found a way that worked. 🤔 A good way to check whether /proc is working correctly is that when you manually run linux-sandbox e.g. with a shell and you look inside /proc, you shouldn't see any processes from outside the sandbox.

Yes it works

src/main/tools/linux-sandbox-pid1.cc

philwo · 2021-06-08T15:37:30Z

src/main/tools/linux-sandbox-pid1.cc

@@ -349,8 +413,13 @@ static void SetupNetworking() {
 }

 static void EnterSandbox() {


WDYT, should we rename this to EnterWorkingDirectory?

I think it makes sense, done

philwo · 2021-06-08T15:40:10Z

src/main/tools/linux-sandbox-pid1.cc

+
+static void CreateEmptyFile() {
+  // This is used as the base for bind mounting.
+  CreateTarget("tmp", true);


Should we check the return value of CreateTarget for errors?

larsrc-google · 2021-06-08T18:26:18Z

@philwo I was actually the one who suggested using the methods from blaze_util instead of adding yet another reimplementation. Seems like that's what a util class is for. Is this util particularly heavy? Should it maybe be split?

philwo · 2021-06-08T18:30:05Z

@larsrc-google But the one method was already there and worked fine and the other two are single lines of code. I think this is taking "DRY" a bit too far.

The sandbox binary is time-tested, performance and correctness critical code, I can't even tell for sure whether this PR will not break something subtle that we will only discover months down the line. This is nothing against this particular code, but more a general cautionary approach I keep around this code - for example, we still found bugs in the signal handling code of the existing code I think just last year, even though it "worked" fine for years already - until it finally didn't under some environmental circumstances.

I would really like to keep it as simple as possible so you can read it in one go and reason about it and remove any abstractions or dependencies on "generic" implementations of functions.

larsrc-google · 2021-06-09T07:40:13Z

@philwo My original comment (https://github.com/bazelbuild/bazel/pull/13279/files/0a9c8f7dc80255e925189c5ac94099fca5a172f3#r608589967) was not intended to apply to existing code so much as to the new implementations being added. Adding new implementations where time-tested ones exist makes it more likely that something will break.

larsrc-google

You seem to have a merge conflict pending.

src/main/java/com/google/devtools/build/lib/actions/FileContentsProxy.java

larsrc-google · 2021-06-08T15:42:23Z

src/main/java/com/google/devtools/build/lib/sandbox/HardlinkedSandboxedSpawn.java

+      Set<Path> writableDirs,
+      TreeDeleter treeDeleter,
+      @Nullable Path statisticsPath,
+      Boolean sandboxDebug) {


Pass in a plain boolean instead of a Boolean object, since you're using it that way anyway.

larsrc-google · 2021-06-08T15:44:15Z

src/main/java/com/google/devtools/build/lib/sandbox/HardlinkedSandboxedSpawn.java

+    hardLinkRecursive(source, target);
+  }
+
+  private void hardLinkRecursive(Path source, Path target) throws IOException {


Please add JavaDoc. In particular explain the failure modes.

src/main/java/com/google/devtools/build/lib/sandbox/HardlinkedSandboxedSpawn.java

larsrc-google · 2021-06-08T15:50:32Z

src/main/java/com/google/devtools/build/lib/sandbox/HardlinkedSandboxedSpawn.java

+        throw new IllegalArgumentException(source + " is a subdirectory of " + target);
+      }
+      target.createDirectory();
+      Collection<Path> entries = source.getDirectoryEntries();


Since you're only using the basename, you can avoid a bunch of string manipulation by using readdir() instead of getDirectoryEntries(), similar to WorkerExecRoot.cleanExisting(). And if you use that, you could have the directory loop be main part of this function and use the file/symlink/dir info in Dirent instead of the possibly expensive is*() calls.

this seemed not so trivial, so I decided not to do anything about this.
I will take a look if you are really concerned by the current implementation of this.

src/main/tools/linux-sandbox-pid1.cc

larsrc-google · 2021-06-10T13:51:31Z

src/main/tools/linux-sandbox-options.cc

+          "  -h  if set, create a chroot in the sandbox directory (-s) and only "
+          " mount whats been specified with -M/-m for improved hermeticity\n"
+          "  -s  The sandbox root directory where the chroot will be created, -W"
+          " should be a folder inside this sandbox root directory\n"


Could you add a check that this holds, probably at the end of ParseOptions? You depend on it further down for a substr operation.

Added it in ParseCommandLine

larsrc-google · 2021-06-10T14:02:15Z

src/main/tools/linux-sandbox-pid1.cc

+}
+
+// Recursively creates the file or directory specified in "path" and its parent
+// directories.


Please add that this sets errno and returns -1 for errors.

larsrc-google · 2021-06-10T14:06:40Z

src/main/tools/linux-sandbox-pid1.cc

+
+  // Create the parent directory.
+  if (CreateTarget(dirname(strdupa(path)), true) < 0) {
+    DIE("CreateTarget");


Adding the name of the directory we failed to create would be helpful here (but easier if CreateTarget dies in all errors).

larsrc-google · 2021-06-10T14:07:15Z

src/main/tools/linux-sandbox-pid1.cc

+// certain filesystems (e.g. XFS).
+static void LinkFile(const char *path) {
+  if (link("tmp/empty_file", path) < 0) {
+    DIE("link");


Please add the file we were trying to link to in the message.

google-cla · 2021-06-14T14:08:35Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

google-cla · 2021-08-10T16:30:04Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

google-cla · 2021-08-10T17:36:25Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

Adds linux-sandbox flag: --experimental_use_hermetic_linux_sandbox - Configure linux-sandbox to run in a chroot environment to prevent access to files not mentioned in the bazel rules unless they can be found via explicitly whitelisted directories using --sandbox_add_mount_pair create hardlinks instead of symlinks, and fallback to copying. In case of writes to input files, the build will be aborted.

frazze-jobb · 2021-08-12T14:00:10Z

Back from vacation now,
I had to reset the branch, sync it and squash all the changes, because the branch had turned real nasty with my and other peoples commits appearing multiple times, hopefully it doesn't ruin the review to much. I no longer have failing tests at least.

I have acted on all your comments @philwo @larsrc-google.
Is it good enough for a merge?

philwo

Thank you. This looks fine to me.

@larsrc-google WDYT?

larsrc-google · 2021-09-01T15:42:27Z

I agree. I have imported it and all tests pass. There are a few style things to fix, I'll take care of that.

ulrfa · 2021-11-08T10:20:23Z

src/main/tools/linux-sandbox-pid1.cc

+  if (CreateTarget("dev", true) < 0) {
+    DIE("CreateTarget /dev");
+  }
+  const char *devs[] = {"/dev/null", "/dev/random", "/dev/urandom", "/dev/zero", NULL};


The sandbox would be less hermetic if /dev/shm is bind mounted, since then actions could communicate with each other via /dev/shm.

Would it make sense to require an explicit --sandbox_add_mount_pair for use cases where /dev/shm is desired to be bind mounted?

Summary: This diff updates our Bazel version to 5.0.0. I've re-worked our patches on top of the upstream 5.0.0 tag. - `linux-sandbox` has changed a bunch (mostly in `linux-sandbox-pid1.cc`) so I had to rewrite a bunch of the patch. The main change is that upstream has added a bunch of logic from our own patch in order to support the new hermetic sandbox flag (see bazelbuild/bazel#13279). So I've cleaned things up, removed some of our code and instead called their new code. One change is that they hardlink an empty file outside of the sandbox rather than creating new files, which sounds ok. Note that we might be able to remove even more of our own patch in favor of their hermetic support but we can do that later. - The merkle tree computation moved from `RemoteSpawnCache` to the `RemoteExecutionService` We should be able to rewrite the patch fairly easily but they've also added an (in-process) cache for those trees (see bazelbuild/bazel#13879) so it might be helping with the slowness that we were seeing before. I'm inclined to not apply the patch to start with and we can add it back if things get much slower. The changes are on the `dbx-2022-02-25-5.0.0` branch in the bazel repo. Here's the list of our own commits on top of upstream: - [[ https://sourcegraph.pp.dropbox.com/bazel/-/commit/5a121a34b1a2a39530bf6cecc3892fc4509a1735?visible=2 | DBX: Helper scripts to build dbx bazel ]] - [[ https://sourcegraph.pp.dropbox.com/bazel/-/commit/c3707dea392806b81f2892d46ede5bf54ef02527?visible=1 | DBX: Point remotejdk URLs to magic mirror ]] - [[ https://sourcegraph.pp.dropbox.com/bazel/-/commit/dc5a85b9a1b710230f2c786fd2cede3adb29370d?visible=2 | DBX: Make sure that the java8 toolchain uses the right options ]] - [[ https://sourcegraph.pp.dropbox.com/bazel/-/commit/497532f9878b3b68582c12766bf034a4de6cc44a?visible=6 | DBX: rootfs patch for the linux-sandbox ]] Also see https://blog.bazel.build/2022/01/19/bazel-5.0.html DTOOLS-1748 Test Plan: Will run the main projects and CI and make sure that things still work. Ran `bzl tool //dropbox/devtools/bazel_metrics/benchmarks --target //services/metaserver edit-refresh` on both this diff and master. On 4.1.0 on master: ``` Running no-op reload 5 times... Finished running no-op reload! The results were: min: 3.01s avg: 3.08s p50: 3.08s max: 3.21s Running modify metaserver/static/js/modules/core/uri.ts 5 times... Finished running modify metaserver/static/js/modules/core/uri.ts! The results were: min: 5.30s avg: 5.78s p50: 5.77s max: 6.59s Running modify metaserver/static/css/legacy_browse.scss 5 times... Finished running modify metaserver/static/css/legacy_browse.scss! The results were: min: 4.46s avg: 4.83s p50: 4.69s max: 5.26s Running add file at metaserver/static/js/modules/core/devbox-benchmark-file-{}.ts 5 times... Finished running add file at metaserver/static/js/modules/core/devbox-benchmark-file-{}.ts! The results were: min: 25.69s avg: 26.21s p50: 26.22s max: 26.89s Running modify metaserver/static/error/maintenance.html 5 times... Finished running modify metaserver/static/error/maintenance.html! The results were: min: 4.75s avg: 4.85s p50: 4.75s max: 5.01s ``` On 5.0.0 ``` Running no-op reload 5 times... Finished running no-op reload! The results were: min: 3.48s avg: 3.69s p50: 3.48s max: 3.90s Running modify metaserver/static/js/modules/core/uri.ts 5 times... Finished running modify metaserver/static/js/modules/core/uri.ts! The results were: min: 5.54s avg: 6.34s p50: 5.54s max: 8.59s Running modify metaserver/static/css/legacy_browse.scss 5 times... Finished running modify metaserver/static/css/legacy_browse.scss! The results were: min: 4.34s avg: 4.75s p50: 5.05s max: 5.46s Running add file at metaserver/static/js/modules/core/devbox-benchmark-file-{}.ts 5 times... Finished running add file at metaserver/static/js/modules/core/devbox-benchmark-file-{}.ts! The results were: min: 25.55s avg: 25.96s p50: 25.64s max: 26.71s Running modify metaserver/static/error/maintenance.html 5 times... Finished running modify metaserver/static/error/maintenance.html! The results were: min: 4.79s avg: 5.33s p50: 5.15s max: 5.84s ``` GitOrigin-RevId: 0f466c5a3bde9ed1157ea936bb70826b58f2fbec

google-cla bot added the cla: no label Mar 30, 2021

frazze-jobb marked this pull request as draft March 30, 2021 07:53

frazze-jobb force-pushed the feature/hermetic_sandbox branch from d3ba024 to 0a9c8f7 Compare March 30, 2021 08:39

frazze-jobb marked this pull request as ready for review March 30, 2021 08:57

aiuto requested review from justinhorvitz and philwo March 31, 2021 03:23

This comment has been minimized.

Sign in to view

google-cla bot added cla: yes and removed cla: no labels Mar 31, 2021

justinhorvitz removed their request for review March 31, 2021 21:05

meisterT requested a review from larsrc-google April 1, 2021 04:59

larsrc-google reviewed Apr 7, 2021

View reviewed changes

src/main/tools/linux-sandbox-options.cc Outdated Show resolved Hide resolved

larsrc-google reviewed Apr 7, 2021

View reviewed changes

src/main/tools/linux-sandbox-pid1.cc Outdated Show resolved Hide resolved

src/main/tools/linux-sandbox-pid1.cc Outdated Show resolved Hide resolved

larsrc-google reviewed Apr 26, 2021

View reviewed changes

janakdr reviewed Apr 29, 2021

View reviewed changes

philwo reviewed Jun 8, 2021

View reviewed changes

larsrc-google reviewed Jun 10, 2021

View reviewed changes

google-cla bot added cla: no and removed cla: yes labels Jun 14, 2021

frazze-jobb force-pushed the feature/hermetic_sandbox branch from 6673bd6 to b82342c Compare June 17, 2021 07:27

google-cla bot added cla: yes and removed cla: no labels Jun 17, 2021

google-cla bot added cla: no and removed cla: yes labels Aug 10, 2021

frazze-jobb force-pushed the feature/hermetic_sandbox branch from 80ba322 to ba775b2 Compare August 10, 2021 18:12

google-cla bot added cla: yes and removed cla: no labels Aug 10, 2021

frazze-jobb force-pushed the feature/hermetic_sandbox branch from ba775b2 to e051b6c Compare August 12, 2021 12:27

philwo approved these changes Aug 26, 2021

View reviewed changes

bazel-io closed this in 11f7d80 Sep 6, 2021

This was referenced Oct 18, 2021

Make the sandboxed file system more strict #7313

Closed

Configurable sandbox root #6994

Closed

ulrfa reviewed Nov 8, 2021

View reviewed changes

cameron-martin mentioned this pull request Feb 14, 2023

native_test and native_binary break shared libraries with remote execution bazelbuild/bazel-skylib#437

Open

lberki mentioned this pull request Oct 30, 2023

Enable --incompatible_sandbox_hermetic_tmp by default #19943

Closed

		// Make sure that the sandbox_root path has no trailing slash.
		if (sandbox_root.back() == '/') {

		@@ -349,8 +413,13 @@ static void SetupNetworking() {
		}

		static void EnterSandbox() {

Implement Hermetic sandbox with support for hardlinks #13279

Implement Hermetic sandbox with support for hardlinks #13279

Conversation

frazze-jobb commented Mar 30, 2021

frazze-jobb commented Mar 30, 2021

google-cla bot commented Mar 30, 2021

What to do if you already signed the CLA

Individual signers

Corporate signers

google-cla bot commented Mar 30, 2021

What to do if you already signed the CLA

Individual signers

Corporate signers

aiuto commented Mar 31, 2021

This comment has been minimized.

frazze-jobb commented Mar 31, 2021

justinhorvitz commented Mar 31, 2021

larsrc-google left a comment

Choose a reason for hiding this comment

larsrc-google commented Apr 9, 2021

frazze-jobb commented Apr 12, 2021 • edited Loading

larsrc-google commented Apr 13, 2021

frazze-jobb commented Apr 20, 2021

larsrc-google commented Apr 20, 2021

frazze-jobb commented Apr 26, 2021

larsrc-google left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

frazze-jobb commented Apr 29, 2021

larsrc-google commented Apr 29, 2021

janakdr left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michajlo commented Apr 29, 2021

philwo left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

larsrc-google commented Jun 8, 2021

philwo commented Jun 8, 2021 • edited Loading

larsrc-google commented Jun 9, 2021

larsrc-google left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

google-cla bot commented Jun 14, 2021

google-cla bot commented Aug 10, 2021

google-cla bot commented Aug 10, 2021

frazze-jobb commented Aug 12, 2021

philwo left a comment

Choose a reason for hiding this comment

larsrc-google commented Sep 1, 2021

Choose a reason for hiding this comment

frazze-jobb commented Apr 12, 2021 •

edited

Loading

philwo commented Jun 8, 2021 •

edited

Loading