Control popups of pinned add-ons are invisible in 140.0.2 + 141.0.b5, OS X 10.7.5

[MacMoinsen]

When clicking a pinned-on add-on, the control box is invisible, although the options in the box do react to roll-over and clicking. 👻

OSX Lion 10.7.5, MacBook 2.1 (mid-2007)

[i3roly]

@Pehrsons i have a question for you. i think the root of the issue i was bugging you about earlier this year and all of these other eccentricities could be related to the failure in loading libnss3.dylib?

i looked at the console for 10.7 and saw a tonne of these errors:

maybe it was "before your time", but if you don't know, would you know who to ask?

https://hg-edge.mozilla.org/mozilla-central/rev/15b761835183778c29def19e8f8136d0b1e95d57

there must be a change in libnss that affects 10.7 but not 10.8. one of the comments said this:

+** Note: This method is disabled on MacOS X 10.7 and iOS version 5.0
+** and will always return SQLITE_MISUSE. On those systems,
+** shared cache mode should be enabled per-database connection via
+** [sqlite3_open_v2()] with [SQLITE_OPEN_SHAREDCACHE].
+**
** This interface is threadsafe on processors where writing a
** 32-bit integer is atomic.
**
** See Also: [SQLite Shared-Cache Mode]
*/
-SQLITE_API int sqlite3_enable_shared_cache(int);
+SQLITE_API int SQLITE_STDCALL sqlite3_enable_shared_cache(int);

i think the lack of libnss3.dylib loading properly would explain the lack of text output on the troubleshooting screen, and some other strange behaviour

[Pehrsons]

On PTO so can't dig around. But Image not found? Is the path wrong? Why does it say .dylib.dylib?

If libnss3 failed to load everywhere there wouldn't be any networking at all, I think? Is the socket process around? Are the errors from content processes?

[i3roly]

seems to be triggered by using the address bar to search/browse, which sorta makes sense since suggest was one of the things throwing the error. but so is uniffi-bindgen-gecko-js, but only when i use the bar. simply clicking/browsing around pages in the contentview doesn't seem to trigger it. maybe it's nothing :/ Thanks, Gagan

…

On Jul 7, 2025, at 9:40 AM, Andreas Pehrson ***@***.***> wrote: Pehrsons left a comment (i3roly/firefox-dynasty#68) On PTO so can't dig around. But Image not found? Is the path wrong? Why does it say .dylib.dylib? If libnss3 failed to load everywhere there wouldn't be any networking at all, I think? Is the socket process around? Are the errors from content processes? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.

[Pehrsons]

It seems some chrome js wants to load libnss3 for sqlite when creating the SuggestStore, which is for address bar suggestions. Does your .app have https://searchfox.org/mozilla-central/source/__GENERATED__/__macosx64__/dist/bin/modules/AppConstants.sys.mjs#202, and is there a file with that name?

[i3roly]

yeah everything there looks fine. and i guess it should be if 10.8+ are working okay

  // URL to the hg revision this was built from (e.g.
  // "https://hg.mozilla.org/mozilla-central/rev/6256ec9113c1")
  // On unofficial builds, this is an empty string.
//@line 193 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"
  SOURCE_REVISION_URL: "",

  HAVE_USR_LIB64_DIR:
//@line 199 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"
    false,
//@line 201 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"

  HAVE_SHELL_SERVICE: true,

  MOZ_CODE_COVERAGE: false,

  TELEMETRY_PING_FORMAT_VERSION: 4,

  ENABLE_WEBDRIVER: true,

  REMOTE_SETTINGS_SERVER_URL:
//@line 214 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"
    "https://firefox.settings.services.mozilla.com/v1",
//@line 216 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"

  REMOTE_SETTINGS_VERIFY_SIGNATURE:
//@line 221 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"
    true,
//@line 223 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"

  REMOTE_SETTINGS_DEFAULT_BUCKET:
//@line 228 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"
    "main",
//@line 230 "$SRCDIR/toolkit/modules/AppConstants.sys.mjs"

i have not the slightest clue why libnss3.dylib is being called with a double suffix. but even if i copied the dylib (so that there's two: libnss3.dylib and libnss3.dylib.dylib) it makes no difference.

very strange lol

[Pehrsons]

Hmm, odd. I do note this re double suffix: https://searchfox.org/mozilla-central/rev/584b344830aa2558985675b99aaadbb915402caa/toolkit/components/uniffi-bindgen-gecko-js/components/generated/RustSuggest.sys.mjs#4792

Might a system trace with Instruments reveal anything as to args and error code for dlopen?

[i3roly]

interesting.

instruments does show something, but it doesn't show libnss3.dylib.dylib.

but this stack seems relevant to the errors above. although it shows libnss3.dylib

however it must fail or not succeed since it goes to a trap after mkstringf?

[RJVB]

i3roly left a comment (i3roly/firefox-dynasty#68) instruments does show something, but it doesn't show libnss3.dylib.dylib.

Is it possible that the double dylib is just a printing error?

but this stack seems relevant to the errors above. although it shows libnss3.dylib

You mean the file shows up as loaded by the application? I never verified: does `dlopen` actually close the file if it can't resolve all references? Or possibly more to the point, does the file get closed and unloaded if an unavailable weak-linked symbol is required? Normally such events cause the process to be terminated, but Firefox tends to hang as a result ...

[i3roly]

any idea @Pehrsons? is it the call to sqlite3_load_extension? is that trace the point of error? i was thinking the answer lay in the trace since you didn't respond, but maybe you have notifications turned off on this thread (which si fair, i get it). gonna try to use sqlite.{c,h} prior to commit https://hg-edge.mozilla.org/mozilla-central/rev/8b744f584b3b1c7825e94735275b3ce6bd031a55 and see what happens will probably fail though Thanks, Gagan

…

On Jul 7, 2025, at 10:16 PM, Andreas Pehrson ***@***.***> wrote: Pehrsons left a comment (i3roly/firefox-dynasty#68) Hmm, odd. I do note this re double suffix: https://searchfox.org/mozilla-central/rev/584b344830aa2558985675b99aaadbb915402caa/toolkit/components/uniffi-bindgen-gecko-js/components/generated/RustSuggest.sys.mjs#4792 Might a system trace with Instruments reveal anything as to args and error code for dlopen? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.

[Pehrsons]

Yeah I'm not sure, this isn't my territory.

Does the dylib in question contain the sqlite symbols, like sqlite3_fts5_init?
Or rephrased, do you have a mozsqlite3.dylib?

[Pehrsons]

Re the trace I was thinking you could trace system calls (checking args and return values) rather than profile. Is it possible?

[i3roly]

negative, no mozsqlite3.dylib:

vms-Mac:~ vm$ find ~/Desktop/Firefox.app/ -name *dylib* /Users/vm/Desktop/Firefox.app//Contents/MacOS/libfreebl3.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libgkcodecs.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/liblgpllibs.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libmozavcodec.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libmozavutil.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libmozglue.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libnss3.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libonnxruntime.dylib /Users/vm/Desktop/Firefox.app//Contents/MacOS/libsoftokn3.dylib /Users/vm/Desktop/Firefox.app//Contents/Resources/gmp-clearkey/0.1/libclearkey.dylib

wrt tracing system calls: i don't know how to use instruments lol. i just did the time profiler and tried finding the relevant stuff in the call tree. i just used system trace and it gave me something similar. i just don't know how to use it to find anything related to "libnss3.dylib.dylib" as that seems to be drawing blanks Thanks, Gagan

…

On Jul 12, 2025, at 6:58 AM, Andreas Pehrson ***@***.***> wrote: Pehrsons left a comment (i3roly/firefox-dynasty#68) Re the trace I was thinking you could trace system calls (checking args and return values) rather than profile. Is it possible? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.

[RJVB]

i3roly left a comment (i3roly/firefox-dynasty#68)

...

> /Users/vm/Desktop/Firefox.app//Contents/MacOS/libnss3.dylib

...

i just don't know how to use it to find anything related to "libnss3.dylib.dylib" as that seems to be drawing blanks

Looks like that is just an error in the error message. Or is libnss3 loaded dynamically, on demand?

[i3roly]

it's possible it's an error in the error message, but why would this not happen on all platforms? it's specific to 10.7 (and probably lower). seems strange. Thanks, Gagan

…

On Jul 12, 2025, at 8:37 AM, René Bertin ***@***.***> wrote: RJVB left a comment (i3roly/firefox-dynasty#68) >i3roly left a comment (i3roly/firefox-dynasty#68) ... >> /Users/vm/Desktop/Firefox.app//Contents/MacOS/libnss3.dylib ... >i just don't know how to use it to find anything related to "libnss3.dylib.dylib" as that seems to be drawing blanks Looks like that is just an error in the error message. Or is libnss3 loaded dynamically, on demand? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.