577744 - Exceptions thrown in NSGetFactory are swallowed

Status: NEW

(Core :: XPCOM, defect)

Description

[Philipp Kewisch [:Fallen] [:📆][:🧩]]

Attachment: Testcase - v1

STR:

1. Install Testcase Extension
2. Start Toolkit Application
3. Visit Error console
4. Execute Components.classes['@mozilla.org/test-exception;1'].createInstance();


Expected:
* Exception is shown both on startup and when creating an instance


Actual:

* This is the only error message you get:

Error: uncaught exception: [Exception... "Component returned failure code: 0x80570015 (NS_ERROR_XPC_CI_RETURNED_FAILURE) [nsIJSCID.createInstance]"  nsresult: "0x80570015 (NS_ERROR_XPC_CI_RETURNED_FAILURE)"  location: "JS frame :: javascript:%20Components.classes['@mozilla.org/test-exception;1'].createInstance(); :: <TOP_LEVEL> :: line 1"  data: no]

Comment 1

[Benjamin Smedberg]

Hrm. I get the following in my error console:

Error: Component {2ece11c6-720b-496c-a047-3fe0372bed4f} is being loaded...
Source File: file:///c:/builds/test-profile/extensions/[e-mail]/components/testException.js
Line: 39
Error: [Exception... "'Loading via category manager works, but now try with:
Components.classes['@mozilla.org/test-exception;1'].createInstance();' when calling method: [xpcIJSGetFactory::get]"  nsresult: "0x8057001e (NS_ERROR_XPC_JS_THREW_STRING)"  location: "<unknown>"  data: no]

These happen during profile-after-change. Then, when I run the .createInstance bit, I get:

Error: Component {2ece11c6-720b-496c-a047-3fe0372bed4f} is being loaded...
Source File: file:///c:/builds/test-profile/extensions/[e-mail]/components/testException.js
Line: 39
Error: uncaught exception: [Exception... "Component returned failure code: 0x80570015 (NS_ERROR_XPC_CI_RETURNED_FAILURE) [nsIJSCID.createInstance]"  nsresult: "0x80570015 (NS_ERROR_XPC_CI_RETURNED_FAILURE)"  location: "JS frame :: javascript:%20Components.classes['@mozilla.org/test-exception;1'].createInstance(); :: <TOP_LEVEL> :: line 1"  data: no]

I wonder if this is because there is JS higher up on the stack.

Comment 2

[Benjamin Smedberg]

In NS_ScriptErrorReporter we're hitting this early-return which bz added in bug 461917.

If I understand everything here correctly, this code is intended to make sure that exceptions which are thrown across XPConnect boundaries are only reported if the caller isn't prepared to handle them. I don't understand why we don't have a stack of "I'm prepared to handle exceptions", rather than assuming that is the case, but oh well for now. I think the best thing to do here is probably something like JS_SaveFrameChain/JS_RestoreFrameChain around the component-eval/NSGetFactory calls. cc'ing bz/mrbkap for thoughts/confirmation.

Comment 3

[Boris Zbarsky [:bzbarsky]]

I can confirm the first half of comment 2....  I have no idea about the second part, offhand. :(