The problem you just found is intimately related to the other one. ECL
uses a garbage collector, evaluation environments and other stuff that
has to be created *for each* thread where you want to execute or
evaluate Common Lisp code. In the simple modification you made the
routine new_Command1 just
On Jul 31, 2010, at 5:49 PM, Juan Jose Garcia-Ripoll wrote:
> Yes, there seems to be one issue remaining. If you start the 4D application
> and attempt to quit (just from the Finder, not using GDB), then 4D hangs. It
> has to be killed using the dock menu or process manager.
>
> Here I cannot
On Sat, Jul 31, 2010 at 8:44 PM, John DeSoi wrote:
> Thanks so much for your help so far. I think we are almost there. I would
> have never figured it out :).
>
I must say I thought all the time that the cause was related to the use of
secondary threads for the plugin: it is the most fragile and
Hi Juanjo,
Thanks so much for your help so far. I think we are almost there. I would have
never figured it out :).
On Jul 31, 2010, at 10:32 AM, Juan Jose Garcia-Ripoll wrote:
> Finally, to shut down the plugin one needs some mechanism to communicate it
> to the thread that it should stop. I
On Sat, Jul 31, 2010 at 4:32 PM, Juan Jose Garcia-Ripoll <
juanjose.garciarip...@googlemail.com> wrote:
> In your case ECL is initialized from a secondary thread, because this is a
> GUI, and the garbage collector fails to recognize the limits of the stack.
> Furthermore, the thread in which it is
I will post what I found because it may be important for other plugin
developers.
The Boehm-Weiser garbage collector assumes that it is initialized from the
main thread of a program. That thread has to live as long as the garbage
collector, and in some platforms it has some special properties, suc
On Thu, Jul 29, 2010 at 2:26 PM, John DeSoi wrote:
> I ran it without GDB and 4D crashes (quits and the OS says that the
> application crashed). Below are the details from Crash Reporter. I called
> GC_init() rather than cl_boot. This is OS X 10.5.8.
>
This obviously points out a problem in the
On Jul 29, 2010, at 4:20 AM, Juan Jose Garcia-Ripoll wrote:
> This is _not_ necessarily a crash. If this is OS X, this signal is used by
> the garbage collector to detect when memory is written to, marking it for
> inspection later on. What I meant is to really run a program, no GDB
> involved
On Thu, Jul 29, 2010 at 2:34 AM, John DeSoi wrote:
>
> On Jul 28, 2010, at 5:36 AM, Juan Jose Garcia-Ripoll wrote:
>
> > There might be some incompatibility between the garbage collector and
> 4D's code. To ensure this is not the case, remove the call to cl_boot() and
> call instead GC_init(). If
On Jul 28, 2010, at 5:36 AM, Juan Jose Garcia-Ripoll wrote:
> There might be some incompatibility between the garbage collector and 4D's
> code. To ensure this is not the case, remove the call to cl_boot() and call
> instead GC_init(). If the collector still crashes then this is the culprit.
I
On Wed, Jul 28, 2010 at 2:13 AM, John DeSoi wrote:
> I'm trying to create a 4D plugin to host ECL (OS X 10.5.8). I call cl_boot
> when the plugin is initialized, but 4D always crashes. If I trace under GDB,
> it appears to crash in GC_push_all_eager at the line "q = (ptr_t)(*p);"
>
There might b
Hi Louis,
That is exactly what I have.
Thanks for the suggestion.
John
On Jul 27, 2010, at 10:55 PM, Louis Höfler wrote:
> I initialize it like this:
> int argc = 1;
> char* argv[1];
> argv[0] = "";
> cl_boot(argc, argv);
>
> Using a unitialized variable for cl_boot will crash it.
John
ecls-list@lists.sourceforge.net
Betreff: [Ecls-list] cl_boot crash
I'm trying to create a 4D plugin to host ECL (OS X 10.5.8). I call cl_boot
when the plugin is initialized, but 4D always crashes. If I trace under GDB,
it appears to crash in GC_push_all_eager at the line "q = (ptr_t)(*p);&qu
I'm trying to create a 4D plugin to host ECL (OS X 10.5.8). I call cl_boot when
the plugin is initialized, but 4D always crashes. If I trace under GDB, it
appears to crash in GC_push_all_eager at the line "q = (ptr_t)(*p);"
I have included part of the crash report below. Unfortunately, I know ju
14 matches
Mail list logo
