[DGD] Error with 1.5.5

[Blain]

The bug is coming from the existence of an undeclared variable.

On Jul 17, 2017 17:34, "Raymond Jennings" <shentino at gmail.com> wrote:

> Actually, if your version successfully called "compile_error", but
> dworkin's aborted due to it being missing, could you double check and make
> sure the version of driver.c that you posted is a complete copy of your
> version?
>
> Dworkin is correct that a call to a non-existant function in the driver
> object is rightly supposed to be a fatal error.
>
> (see http://shentino.github.io/phantasmal/DGD/Running/DGDCrash.html, dgd
> list message dated Fri Oct 24 08:07:01 2003)
>
> However, I did notice that your asserted behavior diverges from dworkin's
> observations.
>
> Also, a compile error in the driver's source should prevent the driver from
> being called at all, and IIRC, the attempt to call "compile_error" would
> itself have triggered a driver object compile (iirc, attempts to call a
> driver function automatically compile the driver), which in turn would
> trigger a runtime error of "compilation within compilation" which
> phantasmal's docs cited as a symptom of an improperly installed/configured
> kernel library (which contains the driver object).
>
> Dworkin is correct that call_out as a kfun requires a minimum of two
> arguments, the name of the function to be called out and the delay for the
> call.  Failing to conform to this is a syntax error (and would have led to
> the aforementioned compile_error call leading to a reentrant compile).
>
> I'm a little surprised that your description differs from dworkin, and in
> theory, unless you've erred in your copypaste, the driver source itself
> should have failed to compile after choking on a call_out based syntax
> error, and the compilation of foo.c would in theory not have happened in
> the first place.
>
> I hope I'm not overthinking this lol.
>
> But at this point I suspect a botched copypaste to the ml for your source
> code.  Could you double check?
>
> On Mon, Jul 17, 2017 at 12:49 AM, Raymond Jennings <shentino at gmail.com>
> wrote:
>
> > If you *were* able to reproduce a true segfault (and not just an abort),
> a
> > stack trace would be helpful.
> >
> > Do make sure it's not due to an infinite recursion though.
> >
> > On Sun, Jul 16, 2017 at 11:24 PM, Felix A. Croes <felix at dworkin.nl>
> wrote:
> >
> >> Blain <blain20 at gmail.com> wrote:
> >>
> >> > I triggered a segfault after compile_error and before runtime_error
> are
> >> > called in the driver object.  I created a bare bones implementation
> and
> >> > determined that the following does bug out.
> >> >
> >> > [/driver.c]
> >> > void initialize() {
> >> >   call_out("load_foo");
> >> > }
> >> >
> >> > void load_foo() {
> >> >   compile_object("/foo");
> >> > }
> >> >
> >> > [/foo.c]
> >> > void foo() {
> >> >   test = ({});
> >> > }
> >> >
> >> > Also, /auto.c and /auto.h are empty files.
> >> >
> >> > With the variable test not being declared in foo.c, DGD segfaults.
> The
> >> > error message given to compile_error is "undeclared variable test".
> >>
> >> The errormessage I get is "/driver.c, 2: too few arguments for function
> >> call_out".
> >>
> >> After fixing that, a fatal error: "missing function in driver object:
> >> compile_error", which does crash DGD, not with a segmentation fault
> >> but with an abort call.  That is proper and expected behaviour.
> >>
> >> Regards,
> >> Felix Croes
> >> ____________________________________________
> >> https://mail.dworkin.nl/mailman/listinfo/dgd
> >
> >
> >
> ____________________________________________
> https://mail.dworkin.nl/mailman/listinfo/dgd