[Edbrowse-dev] out of memory errors

[Edbrowse-dev] out of memory errors

[Chris Brannon]

I'm seeing unhandled out-of-memory conditions in domLink and in
handlerSet.  One of these is eventually ending in a segfault.  I can't
tell which one.

Example from jsdom.cpp, line 1522:
				if (JS_DefineProperty(cw->jss->jcx, owner,
						  idname, vv, NULL, NULL, attr) == JS_FALSE)
return NULL;

The failed domLink calls produce hundreds of "out of memory" messages.
We're doing error checking and returning NULL on error, but
javaSessionFail is never called.
So how do you all want to fix it?  Should I just add javaSessionFail()
to the error legs?  I'm willing to write the patch today.

handlerSet is a little more complicated, because it calls
JS_CompileFunction, which can fail for lots of reasons such as a syntax
error.  We're not doing any error checking there at all, but we should
at least try to catch the out of memory condition.
So I wonder about adding the call to javaSessionFail(); to
my_ErrorReporter directly:

	if (report && report->errorNumber == JSMSG_OUT_OF_MEMORY ||
	    message && strstr(message, "out of memory")) {
		i_puts(MSG_JavaMemError);
		javaSessionFail();
}

-- Chris

Re: [Edbrowse-dev] out of memory errors

[Adam Thompson]

On Sun, Feb 23, 2014 at 10:41:27AM -0800, Chris Brannon wrote:
> I'm seeing unhandled out-of-memory conditions in domLink and in
> handlerSet.  One of these is eventually ending in a segfault.  I can't
> tell which one.
> 
> Example from jsdom.cpp, line 1522:
> 				if (JS_DefineProperty(cw->jss->jcx, owner,
> 						  idname, vv, NULL, NULL, attr) == JS_FALSE)
> return NULL;
> 
> The failed domLink calls produce hundreds of "out of memory" messages.
> We're doing error checking and returning NULL on error, but
> javaSessionFail is never called.

Sorry that was my fault and comes from (I think)
when javaSessionFail destroyed context.
I think the idea at the time was to  put return checks on the
domLink calls, but given the way things work now,
I've just fixed (and pushed the changes to)
domLink to call javaSessionFail and return NULL as I was looking through the 
code  anyway.

> handlerSet is a little more complicated, because it calls
> JS_CompileFunction, which can fail for lots of reasons such as a syntax
> error.  We're not doing any error checking there at all, but we should
> at least try to catch the out of memory condition.

Yeah, we should.
> So I wonder about adding the call to javaSessionFail(); to
> my_ErrorReporter directly:
> 
> 	if (report && report->errorNumber == JSMSG_OUT_OF_MEMORY ||
> 	    message && strstr(message, "out of memory")) {
> 		i_puts(MSG_JavaMemError);
> 		javaSessionFail();
> }

Looks good. Out of interest are there really cases where the error number
doesn't reflect the out of memory condition but the message does?

Cheers,
Adam.