Bug 2068283 - Generic console support by fiji-flo · Pull Request #1339 · mozilla/enterprise-firefox

fiji-flo · 2026-08-25T21:25:00Z

Description

I'll edit this later, opening for sharing early.

Bugzilla: Bug-2068283

Every enterprise deployment currently requires a per-customer repack that bakes the console address into the AutoConfig file (firefox.cfg). This adds a generic build: one artifact that ships without a console address and gets configured at first startup, so the same build can serve any organization (downloads, evaluations, QA).

Implementation Details:

--with-enterprise-console-url still bakes a real address in for repacks. Without it the build is generic and firefox.cfg carries the placeholder FIREFOX_ENTERPRISE_GENERIC
At startup, before profile selection, the placeholder is resolved in this order:
1. the MOZ_ENTERPRISE_CONSOLE_URL environment variable (test harness override),
2. the consoleAddress persisted in felt.json (UAppData) by the setup dialog,
3. otherwise a modal pre-profile setup dialog (chrome://felt/content/consoleSetup.xhtml, FELT UI only; skipped under MOZ_AUTOMATION and in background-task mode) asks for the address, persists it to felt.json, and relaunches so the early consumers (crash reporter ServerURL, update URL, FELT connection) see the configured value from the start.
--reset-console-address clears the stored address so the dialog runs again.
The resolution logic lives once, in the new IO-free enterprise-console crate (toolkit/components/enterprise/rust), shared by all native consumers: browser startup through the felt crate's FFI (XRE_ReadEnterpriseConsoleAddress / XRE_ParseEnterpriseServerURL in toolkit/xre/CreateAppData.cpp), and the standalone crash reporter client. ConsoleClient.sys.mjs mirrors it in JS.
mozrunner and Marionette set MOZ_ENTERPRISE_CONSOLE_URL to a closed local port by default so generic builds never block on the dialog under automation.

Screenshots

Testing

Added tests
Manual testing performed

lissyx · 2026-08-26T11:27:59Z

    NS_ENSURE_SUCCESS(rv, rv);
  }

+  if (serverUrl.EqualsLiteral(ENTERPRISE_CONSOLE_PLACEHOLDER)) {


Out of safety I think this should be reversed: if the value is not the place holder, immediately return NS_ERROR_NOT_AVAILABLE (or maybe NS_ERROR_INVALID_ARG would be more meaningful

That's not the flow here. If the value is not the place holder we need to continue. NS_ERROR_NOT_AVAILABLE is only returned if there's no URL stored / in evn.

lissyx · 2026-08-26T11:33:39Z

+# FIREFOX_ENTERPRISE_GENERIC is a placeholder: on first launch the console
+# setup dialog asks for the address and persists it in felt.json. Repacks
+# replace it with the actual console address.
+enterprise.console.address=FIREFOX_ENTERPRISE_GENERIC


But then we break running local builds without requiring extra step from developpers, which adds friction

I'd rather prefer that we do not change the default distribution.ini to such placeholder. We may at some point want to point it to a different instance than stage though ?

But I'd prefer if we had a dedicated repack that will set the placeholder value

This way, local build are still immediately actionable and the console setup step works like you want from the specific repack

It can also make sure that I cannot inadvertenly have a wrong console in memory of my felt.json storage that gets picked up when running random builds, scoping the feature to those generic repacks

lissyx · 2026-08-26T11:47:00Z

That looks nice!

lissyx · 2026-08-26T11:51:25Z

We need to be extra careful about reviewing further changes using those function, as we may end up in a case where FeltStorage.sys.mjs and those are manipulating the same files at the same time. I think for the time being it's OK, even on macOS where we may have running processes without a window.

But maybe a naïve locking mechanism may be good to protect us for the future?

Yes, we might want to think about this outside of this PR. Multiple Enterprise instances would read/write from that file concurrently right now.

Can we file a follow up for a trivial locking system (e.g. .lock file) at least? To protect us?

Other option, but it would start to get out of scope so a follow up would be nice, would be to convert the FeltStorage code to directly rely on rust so the locking could be done there?

lissyx · 2026-08-26T11:52:54Z

+    if (!(this._feltStorage instanceof lazy.JSONFile)) {
+      this._feltStorage = new lazy.JSONFile({
+        path: this.FELT_FILE_PATH,
+      });
+    }


Should we rather guard with a boolean, e.g. this._initialized ?

Also why do we need this?

I assume it's because we hit now two call sites of FeltStorage.init() one in ConsoleClient and one in consoleSetup.js ?

lissyx · 2026-08-26T12:00:24Z

+  nsAutoString path;
+  rv = file->GetPath(path);
+  NS_ENSURE_SUCCESS(rv, rv);
+  CopyUTF16toUTF8(path, aOutPath);


nsCString is safe for windows paths ? What if the path contains UTF-8 or UTF-16 chars?

Taken from https://searchfox.org/enterprise-main/source/toolkit/xre/nsUpdateDriver.cpp#100

lissyx · 2026-08-26T12:03:20Z

-  XRE_ParseEnterpriseServerURL(*mAppData);
-  // Ignoring nsresult. If console url is not found in a release build, the
-  // default server url is empty and crash reports will fail to submit.
+  if (CheckArg("reset-console-address") == ARG_FOUND) {


So firefox --reset-console-address would allow to trigger again the dialog, but only if its distribution.ini contains the placeholder value.

At my first reading of this part of the code I was afraid, so it may be worth commenting here to make this context crystal clear

Can you suggest an improvement to my comments?

I'll try to come up with something yes

lissyx · 2026-08-26T12:04:36Z

+  // Never block automation on the interactive setup dialog: test harnesses
+  // launch the browser before any console address exists (Marionette's first
+  // launch precedes the test setUp that provides one via distribution.ini or
+  // MOZ_ENTERPRISE_CONSOLE_ADDRESS) and would time out waiting for it.


I think this becomes irrelevant if we move to making such a generic repack instead of changing the default distribution.ini value and the whole checks can be simplified

lissyx · 2026-08-26T12:07:01Z

+  const bool enterpriseConsoleSetupAllowed =
+      !EnvHasValue("MOZ_AUTOMATION") && !CheckArgExists("marionette");
+  if (gEnterpriseConsoleSetupNeeded && enterpriseConsoleSetupAllowed &&
+      is_felt_ui()
+#  ifdef MOZ_BACKGROUNDTASKS
+      && !BackgroundTasks::IsBackgroundTaskMode()
+#  endif
+  ) {
+    rv = ShowEnterpriseConsoleSetup(mNativeApp);
+    if (rv == NS_ERROR_LAUNCHED_CHILD_PROCESS || rv == NS_ERROR_ABORT) {
+      *aExitFlag = true;
+      return 0;
+    }
+    if (NS_FAILED(rv)) {
+      return 1;
+    }
+  }


Implementing via a generic repack instead of changing local builds values, we should be able to simplify to

Suggested change

const bool enterpriseConsoleSetupAllowed =

!EnvHasValue("MOZ_AUTOMATION") && !CheckArgExists("marionette");

if (gEnterpriseConsoleSetupNeeded && enterpriseConsoleSetupAllowed &&

is_felt_ui()

# ifdef MOZ_BACKGROUNDTASKS

&& !BackgroundTasks::IsBackgroundTaskMode()

# endif

) {

rv = ShowEnterpriseConsoleSetup(mNativeApp);

if (rv == NS_ERROR_LAUNCHED_CHILD_PROCESS || rv == NS_ERROR_ABORT) {

*aExitFlag = true;

return 0;

}

if (NS_FAILED(rv)) {

return 1;

}

}

const bool enterpriseConsoleSetupAllowed = is_felt_ui()

# ifdef MOZ_BACKGROUNDTASKS

&& !BackgroundTasks::IsBackgroundTaskMode()

# endif

;

if (gEnterpriseConsoleSetupNeeded && enterpriseConsoleSetupAllowed) {

rv = ShowEnterpriseConsoleSetup(mNativeApp);

if (rv == NS_ERROR_LAUNCHED_CHILD_PROCESS || rv == NS_ERROR_ABORT) {

*aExitFlag = true;

return 0;

}

if (NS_FAILED(rv)) {

return 1;

}

}

I'll look at this part tomorrow.

lissyx · 2026-08-26T12:13:14Z

@fiji-flo Thanks for taking care of this, I think it's overall right, I just really want that we do not change the default build output and rather make such a "generic" repack that will run on CI

lissyx marked this pull request as draft August 26, 2026 11:01

lissyx reviewed Aug 26, 2026

View reviewed changes

Comment thread toolkit/components/enterprise/modules/ConsoleClient.sys.mjs

lissyx reviewed Aug 26, 2026

View reviewed changes

Comment thread toolkit/components/enterprise/modules/ConsoleClient.sys.mjs

lissyx reviewed Aug 26, 2026

View reviewed changes

fiji-flo force-pushed the generic-console branch from 6c4d199 to aecca40 Compare August 31, 2026 16:37

lissyx reviewed Sep 1, 2026

View reviewed changes

Comment thread toolkit/xre/nsAppRunner.cpp

fiji-flo added 7 commits September 1, 2026 19:12

Generic console support

67c2293

fix tests

f3fb9eb

support mozconfig

85d919a

init storage

dd5bf2f

set console address for tests

42bdf1a

Add support to crash reporter and unifiy code in rust

4433a24

fix rebase

2340dbd

fiji-flo force-pushed the generic-console branch from 5268507 to 2340dbd Compare September 1, 2026 18:32

fiji-flo added 2 commits September 1, 2026 23:31

Surface errors on the console address path instead of swallowing them

92534d3

comment

2f738de

fiji-flo changed the title ~~[DRAFT] Generic console support~~ Bug 2068283 - Generic console support Sep 1, 2026

fiji-flo marked this pull request as ready for review September 1, 2026 21:55

fiji-flo added 3 commits September 2, 2026 09:26

https for tests

bff76f7

revert thumbnails test

51add4a

cleanup

37b3edc

fix crash reporter data dir

024290c

Uh oh!

Conversation

fiji-flo commented Aug 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Screenshots

Testing

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lissyx Aug 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

lissyx commented Aug 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lissyx Aug 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lissyx commented Aug 26, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

fiji-flo commented Aug 25, 2026 •

edited

Loading

lissyx Aug 26, 2026 •

edited

Loading

lissyx Aug 31, 2026 •

edited

Loading