I had some segfaults that I suspected where
generated by trying to use objects that
were already collected by the GC.

So I place a gc.FullCollect on my code and now I'm sure.

It's a pitty that to use D now I'm gonna have to
read the pages on GC and memory model and ...
I'm writting higher level apps I don't need to know.

To make D accessible to more users this needs to be corrected.

Anyone wrote something else on the GC
(ok, ok, I'll browse the links...)

We need another D group for applications.

I'm embaraced posting these questions here
and getting no response at all...

Ant

> I had some segfaults that I suspected where
> generated by trying to use objects that
> were already collected by the GC.

That's bad

> So I place a gc.FullCollect on my code and now I'm sure.
>
> It's a pitty that to use D now I'm gonna have to
> read the pages on GC and memory model and ...
> I'm writting higher level apps I don't need to know.

Agreed. I've not yet written sufficiently complex apps such that I can determine whether or not having a mandatory GC is an issue. Nonetheless, if we must have it, then it certainly should work seemlessly, and higher level code should not need to worry.

I would think/hope that this is merely a wrinkle that needs to ironed out of the implementation, rather than a fundamental flaw. Let's hope so, anyway!

> To make D accessible to more users this needs to be corrected.

Indeed. I'd be surprised if Walter doesn't jump all over this asap. He tends not to like these embarassing ones ...

> Anyone wrote something else on the GC
> (ok, ok, I'll browse the links...)
>
> We need another D group for applications.

Agreed. We definitely need more D groups. I'd like

D.lang
D.appl
D.bluesky
D.bugs

> I'm embaraced posting these questions here
> and getting no response at all...

Hopefull this cheerful rejoinder will refresh your spirits, and have you posting with gay abandon!

The newsgroup's gone very quite of late. I've been wondering whether my newsreader had a bug. Maybe everyone's busy with their real jobs ...

> 
> The newsgroup's gone very quite of late. I've been wondering whether my
> newsreader had a bug. Maybe everyone's busy with their real jobs ...
> 

Real jobs suck sometimes.  I'd rather be working on D.  But bills, bills, bills.

In article <bmkua1$1ohe$1@digitaldaemon.com>, Ant says...
>
>I had some segfaults that I suspected where
>generated by trying to use objects that
>were already collected by the GC.

My fault, big bug found.

That's what I expect to hear from the group:
"Can't be - check your code"
Or something like that...

some of my original post is still valid.

Ant

"Brad Anderson" <brad@sankaty.com> wrote in message news:bml90m$26dq$1@digitaldaemon.com...
> >
> > The newsgroup's gone very quite of late. I've been wondering whether my newsreader had a bug. Maybe everyone's busy with their real jobs ...
> >
>
> Real jobs suck sometimes.  I'd rather be working on D.  But bills, bills,
bills.

You should move to Australia. There aren't any jobs to be had, so there's plenty of time for other things. ;)

"Ant" <Ant_member@pathlink.com> wrote in message news:bmle6v$2cqq$1@digitaldaemon.com...
> In article <bmkua1$1ohe$1@digitaldaemon.com>, Ant says...
> >
> >I had some segfaults that I suspected where
> >generated by trying to use objects that
> >were already collected by the GC.
>
> My fault, big bug found.
>
> That's what I expect to hear from the group:
> "Can't be - check your code"
> Or something like that...
>
> some of my original post is still valid.

I'm pretty confident that the gc is solid. It's been hammered pretty good in another project I worked on.

In article <bmoa27$3er$2@digitaldaemon.com>, Walter says...
>
>
>"Ant" <Ant_member@pathlink.com> wrote in message news:bmle6v$2cqq$1@digitaldaemon.com...
>> In article <bmkua1$1ohe$1@digitaldaemon.com>, Ant says...
>> >
>> >I had some segfaults that I suspected where
>> >generated by trying to use objects that
>> >were already collected by the GC.
>>
>> My fault, big bug found.
>>
>> That's what I expect to hear from the group:
>> "Can't be - check your code"
>> Or something like that...
>>
>> some of my original post is still valid.
>
>I'm pretty confident that the gc is solid. It's been hammered pretty good in another project I worked on.

That's good to know, and we suspected that: every thing is looking prety
solid (as in reliable).

(If anybody wants to know my problem was by having references to
objects stored through externel C functions, prety dumb,hey?
I did have a "todo" comment to remove that...
I guess I was surprise the thing worked on the first place)

I do have a public 'thank you' on my D projects web pages,
but somehow I feel I need to thank you again,
if there is any doubt here it is:

thanks for D!

Ant

"Ant" <Ant_member@pathlink.com> wrote in message news:bmp22u$139j$1@digitaldaemon.com...
> I do have a public 'thank you' on my D projects web pages,
> but somehow I feel I need to thank you again,
> if there is any doubt here it is:
>
> thanks for D!

You're welcome! And thanks for putting up the web pages you do, they're a big help.

Were you using linux? I had to recompile libphobos.a in order to get any GC
at all, so maybe that would help.
good luck,
Ben


"Ant" <Ant_member@pathlink.com> wrote in message news:bmkua1$1ohe$1@digitaldaemon.com...
> I had some segfaults that I suspected where
> generated by trying to use objects that
> were already collected by the GC.
>
> So I place a gc.FullCollect on my code and now I'm sure.
>
> It's a pitty that to use D now I'm gonna have to
> read the pages on GC and memory model and ...
> I'm writting higher level apps I don't need to know.
>
> To make D accessible to more users this needs to be corrected.
>
> Anyone wrote something else on the GC
> (ok, ok, I'll browse the links...)
>
> We need another D group for applications.
>
> I'm embaraced posting these questions here
> and getting no response at all...
>
> Ant
>
>

In article <bmpufl$2b1c$1@digitaldaemon.com>, Ben Hinkle says...
>
>Were you using linux? I had to recompile libphobos.a in order to get any GC
>at all, so maybe that would help.
>good luck,
>Ben

linux.

Maybe that's why the error didn't show up in the first place!
Maybe that's why an earlier test creating 20_000 or 500_000 objects
(depending on the object, can't remeber the exact numbers)
would crash an application!

I do recompile phobos for tests sometimes...

Ant