This is a two part post.

First, I wanted to pol how many users out there were developing under OSX? The threads seem to indicated users under windows or Linux, but I've never heard of anybody under OSX. So who has or is developing under OSX? Anybody?

I also wanted to know if you had any "stability" problems. It seems there are strange bugs with the OSX realease every now and then, and those are only the ones we "see". Has anybody tried D with a "relatively" large project? No problems?

The second part of this post is about a specific (OSX related) bug. I don't like drawing attention to it such as this, as I know we are all busy, but I think this one is serious enough to try to push for its investigation: http://d.puremagic.com/issues/show_bug.cgi?id=9720

The reason I'm worried about this bug is that the only condition that seems to trigger it passing an object that has a destructor. I find this is very bothersome, because it can happen with perfectly safe code, and its "observable condition" is perfectly undefined. Finally, it only appears in -O, biting you in the ass in the worst possible moment.

Seeing that this bug exists, I'd have 0 faith to use D under OSX. May I kindly request that those with the required knowledge to fix this bug try to investigate it? Sorry for insisting/asking...

On 2013-03-31 22:02, monarch_dodra wrote:
> This is a two part post.
>
> First, I wanted to pol how many users out there were developing under
> OSX? The threads seem to indicated users under windows or Linux, but
> I've never heard of anybody under OSX. So who has or is developing under
> OSX? Anybody?

I use Mac OS X. I've been using it before DMD was avalible on Mac OS X, used GDC back then. I know there are other developers here that use it as well.

-- 
/Jacob Carlborg

On Sunday, 31 March 2013 at 20:02:40 UTC, monarch_dodra wrote:
> This is a two part post.
>
> First, I wanted to pol how many users out there were developing under OSX? The threads seem to indicated users under windows or Linux, but I've never heard of anybody under OSX. So who has or is developing under OSX? Anybody?
> [snip]

I use osx as well.
most of the things i do are small toy programs.  Though i do play around with vibe a bit and never had any major issue.

I use OSX. I have about 20kloc of D code, and it all works :-)  I have occasionally seen wrong code bugs in the past, but I've reported them all and from what I can remember they have all been fixed.

I can repro your issue. The illegal instruction is a 'popq' instruction called inside insertAfter, just after the destructor for the unnamed 'stuff' is called, before the function returns. I haven't looked into it further than that.

On Sunday, 31 March 2013 at 20:02:40 UTC, monarch_dodra wrote:
> This is a two part post.
>
> First, I wanted to pol how many users out there were developing under OSX? The threads seem to indicated users under windows or Linux, but I've never heard of anybody under OSX. So who has or is developing under OSX? Anybody?
>
> I also wanted to know if you had any "stability" problems. It seems there are strange bugs with the OSX realease every now and then, and those are only the ones we "see". Has anybody tried D with a "relatively" large project? No problems?
>
> The second part of this post is about a specific (OSX related) bug. I don't like drawing attention to it such as this, as I know we are all busy, but I think this one is serious enough to try to push for its investigation: http://d.puremagic.com/issues/show_bug.cgi?id=9720
>
> The reason I'm worried about this bug is that the only condition that seems to trigger it passing an object that has a destructor. I find this is very bothersome, because it can happen with perfectly safe code, and its "observable condition" is perfectly undefined. Finally, it only appears in -O, biting you in the ass in the worst possible moment.
>
> Seeing that this bug exists, I'd have 0 faith to use D under OSX. May I kindly request that those with the required knowledge to fix this bug try to investigate it? Sorry for insisting/asking...

We use D at work, since pre-1, and right now we are using it in Windows, Linux mainly and OSX: we have a discrete codebase, and we are using it for production code in real products.

Actually on OSX I'm using LDC, because simply I can't use objc bindings with DMD (random stack corruptions), and I'm a little too under time pressure to investigate the issue.

I can confirm that we have some problems with -O code: actually we are avoiding it, and we are simply stuck with no optimisations at all...

/Paolo

On 2013-04-01 11:42, Paolo Invernizzi wrote:

> Actually on OSX I'm using LDC, because simply I can't use objc bindings
> with DMD (random stack corruptions), and I'm a little too under time
> pressure to investigate the issue.

What kind of objc bindings are you using? Own implementation or some library I might know about?

-- 
/Jacob Carlborg

On Monday, 1 April 2013 at 10:32:09 UTC, Jacob Carlborg wrote:
> On 2013-04-01 11:42, Paolo Invernizzi wrote:
>
>> Actually on OSX I'm using LDC, because simply I can't use objc bindings
>> with DMD (random stack corruptions), and I'm a little too under time
>> pressure to investigate the issue.
>
> What kind of objc bindings are you using? Own implementation or some library I might know about?

They are a simple declarations of the main objc functions and types made by ourself: nothing really special.

On 2013-04-01 15:27, Paolo Invernizzi wrote:

> They are a simple declarations of the main objc functions and types made
> by ourself: nothing really special.

Ok, I see.

-- 
/Jacob Carlborg

01.04.2013 0:02, monarch_dodra пишет:
> This is a two part post.
>
> The reason I'm worried about this bug is that the only condition that
> seems to trigger it passing an object that has a destructor. I find this
> is very bothersome, because it can happen with perfectly safe code, and
> its "observable condition" is perfectly undefined. Finally, it only
> appears in -O, biting you in the ass in the worst possible moment.

Nothing personal, I do respect you, but

Bitch please!

dmd has the ability to produce broken object files from time to time on every platform [1] so "the worst possible moment" can punish you even when e.g. you remove tracing code or update druntime/Phobos. And yes, it happens often - it hit me on two different projects (guess, how many compile-to-exe non-toy projects I had last time?).

Also OPTLINK tends to stop linking (yes, randomly) when your project grows in size and you use templates (read "ranges") forcing you to do further development without debug info (really nice to read you own code only in asm, isn't it?) thus making D unusable for non-toy Windows development. I have never understand people saying "it's a bullshit, you are just angry, D is usable" as they never ever argue how you can use D (my advice for them to argue me here: "D is usable if you carefully (read "don't") use templates").

Also one have to use contracts and closures very carefully because of tons of wrong-code bugs.

[1] http://d.puremagic.com/issues/show_bug.cgi?id=9044
[2] http://d.puremagic.com/issues/show_bug.cgi?id=6144


P.S.
Still think you "wrong-code with -O switch" issue is serious? I have never think D is nearly ready to pass "-O" to compiler (afraid of CPU explosion on compiled program launch).


P.P.S.
Create a time machine and move back in early 2012 (or just imagine there is no Kenji Hara among us). What do you see? Right, wrong code on almost every `std.algorithm` usage (read "when predicate uses outer scope"). So current situation is still the same heaven it was that time Kenji fixes that. Yes, I remember I was so happy...

-- 
Денис В. Шеломовский
Denis V. Shelomovskij

On 4/1/13 11:21 AM, Denis Shelomovskij wrote:
> 01.04.2013 0:02, monarch_dodra пишет:
>> This is a two part post.
>>
>> The reason I'm worried about this bug is that the only condition that
>> seems to trigger it passing an object that has a destructor. I find this
>> is very bothersome, because it can happen with perfectly safe code, and
>> its "observable condition" is perfectly undefined. Finally, it only
>> appears in -O, biting you in the ass in the worst possible moment.
>
> Nothing personal, I do respect you, but
>
> Bitch please!
[snip]

Denis, the above (as well as most of the message that follows) is entirely inappropriate. It's also compounded by your antics on github, where your contributions are marred by a tendency to bully other contributors and to convert most every disagreement into strife. I have repeatedly asked you kindly to correct that behavior to no effect.

Remember, nobody's holding a gun to your head; the door is always open for entering as well as leaving. It is entirely understandable if you find D unfit for whatever you do, or if its development process is not to your satisfaction. But this ongoing attitude of playing the victim, interpreting the team's shortcomings as incompetence doubled by malice, and picking fights left and right, is not helping you or anybody.

I compel you again to revise your attitude toward the people you work and interact with in the forum and on github. If that does not happen, I will propose to the core team that your github account is banned from our project. Sorry it had to come to this.


Thanks,

Andrei