gnustep-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: OpenBSD: Services problems


From: Fred Kiefer
Subject: Re: OpenBSD: Services problems
Date: Mon, 12 Sep 2011 10:32:43 +0200
User-agent: Mozilla/5.0 (X11; U; Linux x86_64; de; rv:1.9.2.22) Gecko/20110907 SUSE/3.1.14 Thunderbird/3.1.14

On 11.09.2011 14:03, Riccardo Mottola wrote:
On the infamous OpenBSD/sparc computer I get now.

2011-09-11 13:58:38.068 Terminal[24493] NSConnection.m:2601 Assertion
failed in -[NSConnection(Private) _service:forwardForProxy:]. Invalid
parameter not satisfying: type
2011-09-11 13:59:14.192 Vespucci[31124] NSConnection.m:2601 Assertion
failed in -[NSConnection(Private) _service:forwardForProxy:]. Invalid
parameter not satisfying: type

Could you please explain in details what you did to trigger this error? Are these messages triggered when you tried to use service menu items from GWorkspace or did you get them from the specific applications (Terminal, Vespucci) directly and if so, what where you doing there?

The basic problem reported here is that a method signature doesn't have type information attached to it. This gets detected in line 2601 of the file NSConnection.m. There we also have some addition error reporting in place. Could you please run your service applications with --GNU-Debug=dflt. It would also be interesting to get even more information from this method. What object are we trying to access (I think it should be the GSServiceManager) and is the service method found on the object? Of course you should recompile Terminal and Vespucci first, just to make sure they get linked against the current version of base.

I don't quite understand why the code in NSConnection does all the type extraction by itself. To me it looks like we should reuse the code from [NSObject -methodSignatureForSelector:] here and get the type information from the NSMethodSignature. I think the main difference will be in the case where the local object wont implement the method in question. But exactly in that case it would be better to ask the object for the method signature then to invent the type information locally.

David, Richard, you two should understand a lot more about this than me. Is this reasoning correct and if so, do you think it is safe to replace the code that close to release?




reply via email to

[Prev in Thread] Current Thread [Next in Thread]