www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Shared Delegates

reply Andrew Wiley <wiley.andrew.j gmail.com> writes:
Okay, I realize there have been some discussions about this, but I have a
few questions about shared delegates because right now they are definitely
broken, but I'm not sure how.
Take this code example:

synchronized class Thing {
        void doSomeWork(void delegate() work) {
                work();
        }
        void work() {}
}

void main() {
        auto th = new Thing();
        th.doSomeWork(&th.work);
}

This doesn't compile because the type of "&th.work" is "void delegate()
shared", which cannot be cast implicitly to "void delegate()".
My first question would be whether that type is correct. It's true that the
data pointer of the delegate points to a shared object, but given that the
function locks it, does that really matter in this case? I guess I'm just
not clear on the exact meaning of "shared" in general, but it seems like
whether the data is shared or not is irrelevant when the delegate points to
a public member of a synchronized class. If it was a delegate pointing to a
private/protected member (which should be illegal in this case), that would
not be true.
If that type is correct, the problem is that "void delegate() shared"
doesn't parse as a type (there is a workaround because you can create
variables of this type through alias and typeof).

What, exactly, is wrong here?
Oct 17 2011
parent reply Michel Fortin <michel.fortin michelf.com> writes:
On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com> said:

 
 Okay, I realize there have been some discussions about this, but I have a
 few questions about shared delegates because right now they are definitely
 broken, but I'm not sure how.
 Take this code example:
 
 synchronized class Thing {
         void doSomeWork(void delegate() work) {
                 work();
         }
         void work() {}
 }
 
 void main() {
         auto th = new Thing();
         th.doSomeWork(&th.work);
 }
 
 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true that the
 data pointer of the delegate points to a shared object, but given that the
 function locks it, does that really matter in this case? I guess I'm just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate points to
 a public member of a synchronized class. If it was a delegate pointing to a
 private/protected member (which should be illegal in this case), that would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).
 
 What, exactly, is wrong here?
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way. -- Michel Fortin michel.fortin michelf.com http://michelf.com/
Oct 17 2011
next sibling parent reply Benjamin Thaut <code benjamin-thaut.de> writes:
Am 17.10.2011 22:43, schrieb Michel Fortin:
 On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com> said:

 Okay, I realize there have been some discussions about this, but I have a
 few questions about shared delegates because right now they are
 definitely
 broken, but I'm not sure how.
 Take this code example:

 synchronized class Thing {
 void doSomeWork(void delegate() work) {
 work();
 }
 void work() {}
 }

 void main() {
 auto th = new Thing();
 th.doSomeWork(&th.work);
 }

 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true
 that the
 data pointer of the delegate points to a shared object, but given that
 the
 function locks it, does that really matter in this case? I guess I'm just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate
 points to
 a public member of a synchronized class. If it was a delegate pointing
 to a
 private/protected member (which should be illegal in this case), that
 would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).

 What, exactly, is wrong here?
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way.
I reported this exact issue already a few months ago and simply didn't get any comment on it. If you really try to use shared to are going to hit more such problems. See the shared section of my blogpost: http://3d.benjamin-thaut.de/?p=18 -- Kind Regards Benjamin Thaut
Oct 17 2011
parent reply Andrew Wiley <wiley.andrew.j gmail.com> writes:
On Tue, Oct 18, 2011 at 12:53 AM, Benjamin Thaut <code benjamin-thaut.de>wrote:

 Am 17.10.2011 22:43, schrieb Michel Fortin:

  On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:


 Okay, I realize there have been some discussions about this, but I have a
 few questions about shared delegates because right now they are
 definitely
 broken, but I'm not sure how.
 Take this code example:

 synchronized class Thing {
 void doSomeWork(void delegate() work) {
 work();
 }
 void work() {}
 }

 void main() {
 auto th = new Thing();
 th.doSomeWork(&th.work);
 }

 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true
 that the
 data pointer of the delegate points to a shared object, but given that
 the
 function locks it, does that really matter in this case? I guess I'm just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate
 points to
 a public member of a synchronized class. If it was a delegate pointing
 to a
 private/protected member (which should be illegal in this case), that
 would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).

 What, exactly, is wrong here?
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way.
I reported this exact issue already a few months ago and simply didn't get any comment on it. If you really try to use shared to are going to hit more such problems. See the shared section of my blogpost: http://3d.benjamin-thaut.de/?**p=18<http://3d.benjamin-thaut.de/?p=18>
Ah, I was looking through the bug reports and didn't see this exact bug. Did I just fail at searching, or should I file it?
Oct 17 2011
parent reply Benjamin Thaut <code benjamin-thaut.de> writes:
Am 18.10.2011 08:03, schrieb Andrew Wiley:
 On Tue, Oct 18, 2011 at 12:53 AM, Benjamin Thaut <code benjamin-thaut.de
 <mailto:code benjamin-thaut.de>> wrote:

     Am 17.10.2011 22:43, schrieb Michel Fortin:

         On 2011-10-17 20:33:59 +0000, Andrew Wiley
         <wiley.andrew.j gmail.com <mailto:wiley.andrew.j gmail.com>> said:


             Okay, I realize there have been some discussions about this,
             but I have a
             few questions about shared delegates because right now they are
             definitely
             broken, but I'm not sure how.
             Take this code example:

             synchronized class Thing {
             void doSomeWork(void delegate() work) {
             work();
             }
             void work() {}
             }

             void main() {
             auto th = new Thing();
             th.doSomeWork(&th.work);
             }

             This doesn't compile because the type of "&th.work" is "void
             delegate()
             shared", which cannot be cast implicitly to "void delegate()".
             My first question would be whether that type is correct.
             It's true
             that the
             data pointer of the delegate points to a shared object, but
             given that
             the
             function locks it, does that really matter in this case? I
             guess I'm just
             not clear on the exact meaning of "shared" in general, but
             it seems like
             whether the data is shared or not is irrelevant when the
             delegate
             points to
             a public member of a synchronized class. If it was a
             delegate pointing
             to a
             private/protected member (which should be illegal in this
             case), that
             would
             not be true.
             If that type is correct, the problem is that "void
             delegate() shared"
             doesn't parse as a type (there is a workaround because you
             can create
             variables of this type through alias and typeof).

             What, exactly, is wrong here?


         I think what's wrong is that a shared delegate should implicitly
         convert
         to a non-shared one. The delegate is shared since it can be called
         safely from any thread, and making it non-shared only prevent
         you from
         propagating it to more thread so it's not harmful in any way.


     I reported this exact issue already a few months ago and simply
     didn't get any comment on it. If you really try to use shared to are
     going to hit more such problems.

     See the shared section of my blogpost:
     http://3d.benjamin-thaut.de/?__p=18 <http://3d.benjamin-thaut.de/?p=18>


 Ah, I was looking through the bug reports and didn't see this exact bug.
 Did I just fail at searching, or should I file it?
I didn't file it yet, so file it. I do however think that currently there is no intention in changing the way shared works. -- Kind Regards Benjamin Thaut
Oct 17 2011
next sibling parent Andrew Wiley <wiley.andrew.j gmail.com> writes:
On Tue, Oct 18, 2011 at 1:11 AM, Benjamin Thaut <code benjamin-thaut.de>wrote:

 Am 18.10.2011 08:03, schrieb Andrew Wiley:

 On Tue, Oct 18, 2011 at 12:53 AM, Benjamin Thaut <code benjamin-thaut.de
 <mailto:code benjamin-thaut.de**>> wrote:

    Am 17.10.2011 22:43, schrieb Michel Fortin:

        On 2011-10-17 20:33:59 +0000, Andrew Wiley
        <wiley.andrew.j gmail.com
<mailto:wiley.andrew.j gmail.**com<wiley.andrew.j gmail.com>>>
 said:


            Okay, I realize there have been some discussions about this,
            but I have a
            few questions about shared delegates because right now they are
            definitely
            broken, but I'm not sure how.
            Take this code example:

            synchronized class Thing {
            void doSomeWork(void delegate() work) {
            work();
            }
            void work() {}
            }

            void main() {
            auto th = new Thing();
            th.doSomeWork(&th.work);
            }

            This doesn't compile because the type of "&th.work" is "void
            delegate()
            shared", which cannot be cast implicitly to "void delegate()".
            My first question would be whether that type is correct.
            It's true
            that the
            data pointer of the delegate points to a shared object, but
            given that
            the
            function locks it, does that really matter in this case? I
            guess I'm just
            not clear on the exact meaning of "shared" in general, but
            it seems like
            whether the data is shared or not is irrelevant when the
            delegate
            points to
            a public member of a synchronized class. If it was a
            delegate pointing
            to a
            private/protected member (which should be illegal in this
            case), that
            would
            not be true.
            If that type is correct, the problem is that "void
            delegate() shared"
            doesn't parse as a type (there is a workaround because you
            can create
            variables of this type through alias and typeof).

            What, exactly, is wrong here?


        I think what's wrong is that a shared delegate should implicitly
        convert
        to a non-shared one. The delegate is shared since it can be called
        safely from any thread, and making it non-shared only prevent
        you from
        propagating it to more thread so it's not harmful in any way.


    I reported this exact issue already a few months ago and simply
    didn't get any comment on it. If you really try to use shared to are
    going to hit more such problems.

    See the shared section of my blogpost:
    http://3d.benjamin-thaut.de/?_**_p=18<http://3d.benjamin-thaut.de/?__p=18><
 http://3d.benjamin-thaut.de/?**p=18 <http://3d.benjamin-thaut.de/?p=18>>



 Ah, I was looking through the bug reports and didn't see this exact bug.
 Did I just fail at searching, or should I file it?
I didn't file it yet, so file it. I do however think that currently there is no intention in changing the way shared works.
This much is clearly a bug, and Michel's explanation of how shared delegates should work makes a lot of sense. As for the synchronized classes that synchronize access to not-really-shared members, well, I don't think that's changing. The problem is that the compiler can't *guarantee* that the reference you hold is the only reference. TDPL has a discussion of why an "A owns B, so A's lock should be good enough for B" wasn't implemented. As for the overloading on shared, the idea there was that you should really either be multithreaded or not - trying to implement both shared and non-shared versions of code is generally a bad idea and quite bug prone. If it *might* be shared, just synchronize it. Premature optimization is the root of quite a bit of evil, and trying to make some things thread safe and some things not is walking a tightrope without any sense of balance - you won't know when you fall, you'll just see the corruption when you hit the floor.
Oct 17 2011
prev sibling parent Andrew Wiley <wiley.andrew.j gmail.com> writes:
On Tue, Oct 18, 2011 at 1:48 AM, Andrew Wiley <wiley.andrew.j gmail.com>wrote:

 On Tue, Oct 18, 2011 at 1:11 AM, Benjamin Thaut <code benjamin-thaut.de>wrote:

 Am 18.10.2011 08:03, schrieb Andrew Wiley:

 On Tue, Oct 18, 2011 at 12:53 AM, Benjamin Thaut <code benjamin-thaut.de
 <mailto:code benjamin-thaut.de**>> wrote:

    Am 17.10.2011 22:43, schrieb Michel Fortin:

        On 2011-10-17 20:33:59 +0000, Andrew Wiley
        <wiley.andrew.j gmail.com
<mailto:wiley.andrew.j gmail.**com<wiley.andrew.j gmail.com>>>
 said:


            Okay, I realize there have been some discussions about this,
            but I have a
            few questions about shared delegates because right now they
 are
            definitely
            broken, but I'm not sure how.
            Take this code example:

            synchronized class Thing {
            void doSomeWork(void delegate() work) {
            work();
            }
            void work() {}
            }

            void main() {
            auto th = new Thing();
            th.doSomeWork(&th.work);
            }

            This doesn't compile because the type of "&th.work" is "void
            delegate()
            shared", which cannot be cast implicitly to "void delegate()".
            My first question would be whether that type is correct.
            It's true
            that the
            data pointer of the delegate points to a shared object, but
            given that
            the
            function locks it, does that really matter in this case? I
            guess I'm just
            not clear on the exact meaning of "shared" in general, but
            it seems like
            whether the data is shared or not is irrelevant when the
            delegate
            points to
            a public member of a synchronized class. If it was a
            delegate pointing
            to a
            private/protected member (which should be illegal in this
            case), that
            would
            not be true.
            If that type is correct, the problem is that "void
            delegate() shared"
            doesn't parse as a type (there is a workaround because you
            can create
            variables of this type through alias and typeof).

            What, exactly, is wrong here?


        I think what's wrong is that a shared delegate should implicitly
        convert
        to a non-shared one. The delegate is shared since it can be called
        safely from any thread, and making it non-shared only prevent
        you from
        propagating it to more thread so it's not harmful in any way.


    I reported this exact issue already a few months ago and simply
    didn't get any comment on it. If you really try to use shared to are
    going to hit more such problems.

    See the shared section of my blogpost:
    http://3d.benjamin-thaut.de/?_**_p=18<http://3d.benjamin-thaut.de/?__p=18><
 http://3d.benjamin-thaut.de/?**p=18 <http://3d.benjamin-thaut.de/?p=18>>



 Ah, I was looking through the bug reports and didn't see this exact bug.
 Did I just fail at searching, or should I file it?
I didn't file it yet, so file it. I do however think that currently there is no intention in changing the way shared works.
This much is clearly a bug, and Michel's explanation of how shared delegates should work makes a lot of sense.
Filed as http://d.puremagic.com/issues/show_bug.cgi?id=6823 For now, I can just cast things at the call site.
Oct 18 2011
prev sibling parent reply Andrew Wiley <wiley.andrew.j gmail.com> writes:
On Mon, Oct 17, 2011 at 3:43 PM, Michel Fortin <michel.fortin michelf.com>wrote:

 On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:


 Okay, I realize there have been some discussions about this, but I have a
 few questions about shared delegates because right now they are definitely
 broken, but I'm not sure how.
 Take this code example:

 synchronized class Thing {
        void doSomeWork(void delegate() work) {
                work();
        }
        void work() {}
 }

 void main() {
        auto th = new Thing();
        th.doSomeWork(&th.work);
 }

 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true that
 the
 data pointer of the delegate points to a shared object, but given that the
 function locks it, does that really matter in this case? I guess I'm just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate points
 to
 a public member of a synchronized class. If it was a delegate pointing to
 a
 private/protected member (which should be illegal in this case), that
 would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).

 What, exactly, is wrong here?
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way.
Actually, I've been thinking about this some more, and I think that the delegate should only implicitly convert if the argument types are safe to share across threads as well. If I had a class that looked like this: synchronized class Thing2 { void doSomeWork(int i) {} void doSomeOtherWork(Thing2 t) {} void work() {} } The actual argument type in doSomeOtherWork is required to be shared(Thing2) (which isn't a problem here because Thing2 is a synchronized class, but you see the point). These same rules should apply to shared delegates. I'm not sure exactly how this works for value types because they're always safe to pass as arguments yet shared(int) clearly isn't the same as int, but the rules will be the same for delegates as for member functions. With that in mind, I'm not sure exactly how the implicit conversions should work, but the argument types should always be properly share-able.
Oct 19 2011
parent reply Michel Fortin <michel.fortin michelf.com> writes:
On 2011-10-19 20:36:37 +0000, Andrew Wiley <wiley.andrew.j gmail.com> said:

 
 On Mon, Oct 17, 2011 at 3:43 PM, Michel Fortin 
 <michel.fortin michelf.com>wrote:
 
 On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:
 
 
 Okay, I realize there have been some discussions about this, but I have a
 few questions about shared delegates because right now they are definitely
 broken, but I'm not sure how.
 Take this code example:
 
 synchronized class Thing {
 void doSomeWork(void delegate() work) {
 work();
 }
 void work() {}
 }
 
 void main() {
 auto th = new Thing();
 th.doSomeWork(&th.work);
 }
 
 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true that
 the
 data pointer of the delegate points to a shared object, but given that the
 function locks it, does that really matter in this case? I guess I'm just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate points
 to
 a public member of a synchronized class. If it was a delegate pointing to
 a
 private/protected member (which should be illegal in this case), that
 would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).
 
 What, exactly, is wrong here?
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way.
Actually, I've been thinking about this some more, and I think that the delegate should only implicitly convert if the argument types are safe to share across threads as well.
I disagree.
 If I had a class that looked like this:
 synchronized class Thing2 {
        void doSomeWork(int i) {}
        void doSomeOtherWork(Thing2 t) {}
        void work() {}
 }
 
 The actual argument type in doSomeOtherWork is required to be shared(Thing2)
 (which isn't a problem here because Thing2 is a synchronized class, but you
 see the point).
Is it? Whether the argument was shared or not, the thread in which the function code runs can only access thread-local data from that thread, including the arguments and global variables. It won't be able to send references to non-shared data to other threads just because its context pointer is shared (or synchronized). -- Michel Fortin michel.fortin michelf.com http://michelf.com/
Oct 19 2011
parent reply Andrew Wiley <wiley.andrew.j gmail.com> writes:
On Wed, Oct 19, 2011 at 4:11 PM, Michel Fortin <michel.fortin michelf.com>wrote:

 On 2011-10-19 20:36:37 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:


 On Mon, Oct 17, 2011 at 3:43 PM, Michel Fortin <michel.fortin michelf.com
**wrote:
On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:


  Okay, I realize there have been some discussions about this, but I have
 a
 few questions about shared delegates because right now they are
 definitely
 broken, but I'm not sure how.
 Take this code example:

 synchronized class Thing {
 void doSomeWork(void delegate() work) {
 work();
 }
 void work() {}
 }

 void main() {
 auto th = new Thing();
 th.doSomeWork(&th.work);
 }

 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true that
 the
 data pointer of the delegate points to a shared object, but given that
 the
 function locks it, does that really matter in this case? I guess I'm
 just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate points
 to
 a public member of a synchronized class. If it was a delegate pointing
 to
 a
 private/protected member (which should be illegal in this case), that
 would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).

 What, exactly, is wrong here?
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way.
Actually, I've been thinking about this some more, and I think that the delegate should only implicitly convert if the argument types are safe to share across threads as well.
I disagree. If I had a class that looked like this:
 synchronized class Thing2 {
       void doSomeWork(int i) {}
       void doSomeOtherWork(Thing2 t) {}
       void work() {}
 }

 The actual argument type in doSomeOtherWork is required to be
 shared(Thing2)
 (which isn't a problem here because Thing2 is a synchronized class, but
 you
 see the point).
Is it? Whether the argument was shared or not, the thread in which the function code runs can only access thread-local data from that thread, including the arguments and global variables. It won't be able to send references to non-shared data to other threads just because its context pointer is shared (or synchronized).
The problem is that what's behind the context pointer is also shared. If this delegate is just a closure, that doesn't matter, since the context is basically immutable. The problem I see is when the delegate is actually a member function that stores data in an object. If it was passed a reference to non-shared data, it could store that reference in a shared object, breaking transitive shared.
Oct 19 2011
parent reply Michel Fortin <michel.fortin michelf.com> writes:
On 2011-10-19 21:53:12 +0000, Andrew Wiley <wiley.andrew.j gmail.com> said:

 
 On Wed, Oct 19, 2011 at 4:11 PM, Michel Fortin 
 <michel.fortin michelf.com>wrote:
 
 On 2011-10-19 20:36:37 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:
 
 
 On Mon, Oct 17, 2011 at 3:43 PM, Michel Fortin <michel.fortin michelf.com
 **wrote:
On 2011-10-17 20:33:59 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:
 
 
 Okay, I realize there have been some discussions about this, but I have
 a
 few questions about shared delegates because right now they are
 definitely
 broken, but I'm not sure how.
 Take this code example:
 
 synchronized class Thing {
 void doSomeWork(void delegate() work) {
 work();
 }
 void work() {}
 }
 
 void main() {
 auto th = new Thing();
 th.doSomeWork(&th.work);
 }
 
 This doesn't compile because the type of "&th.work" is "void delegate()
 shared", which cannot be cast implicitly to "void delegate()".
 My first question would be whether that type is correct. It's true that
 the
 data pointer of the delegate points to a shared object, but given that
 the
 function locks it, does that really matter in this case? I guess I'm
 just
 not clear on the exact meaning of "shared" in general, but it seems like
 whether the data is shared or not is irrelevant when the delegate points
 to
 a public member of a synchronized class. If it was a delegate pointing
 to
 a
 private/protected member (which should be illegal in this case), that
 would
 not be true.
 If that type is correct, the problem is that "void delegate() shared"
 doesn't parse as a type (there is a workaround because you can create
 variables of this type through alias and typeof).
 
 What, exactly, is wrong here?
 
I think what's wrong is that a shared delegate should implicitly convert to a non-shared one. The delegate is shared since it can be called safely from any thread, and making it non-shared only prevent you from propagating it to more thread so it's not harmful in any way.
Actually, I've been thinking about this some more, and I think that the delegate should only implicitly convert if the argument types are safe to share across threads as well.
I disagree. If I had a class that looked like this:
 synchronized class Thing2 {
 void doSomeWork(int i) {}
 void doSomeOtherWork(Thing2 t) {}
 void work() {}
 }
 
 The actual argument type in doSomeOtherWork is required to be
 shared(Thing2)
 (which isn't a problem here because Thing2 is a synchronized class, but
 you
 see the point).
Is it? Whether the argument was shared or not, the thread in which the function code runs can only access thread-local data from that thread, including the arguments and global variables. It won't be able to send references to non-shared data to other threads just because its context pointer is shared (or synchronized).
The problem is that what's behind the context pointer is also shared. If this delegate is just a closure, that doesn't matter, since the context is basically immutable. The problem I see is when the delegate is actually a member function that stores data in an object. If it was passed a reference to non-shared data, it could store that reference in a shared object, breaking transitive shared.
I think I see what you mean, but you're probably missing one little detail… Implicitly casting the delegate to non-shared won't change what compiled code the delegate points to. The compiled code of the delegate was compiled with a shared 'this' pointer in mind. Since the attributes of the 'this' pointer is engraved in the compiled code when the function was compiled, changing the attributes of the delegate (a mere pointer to that code) won't change how the function treats 'this'. And since the 'this' pointer is shared inside the function you can't assign anything to it that isn't shared. -- Michel Fortin michel.fortin michelf.com http://michelf.com/
Oct 19 2011
parent reply "Robert Jacques" <sandford jhu.edu> writes:
On Wed, 19 Oct 2011 20:51:25 -0400, Michel Fortin <michel.fortin michelf.com>
wrote:
 On 2011-10-19 21:53:12 +0000, Andrew Wiley <wiley.andrew.j gmail.com> said:
[snip] Also, I how shared works is being misunderstood. Making a class synchronized should limit all member functions and field to being shared or immutable, but it doesn't place any limits on the argument to the member functions. So in the below: synchronized class Thing2 { void doSomeWork(int i) {} void doSomeOtherWork(Thing2 t) {} void work() {} } i is of type int, not shared(int).
Oct 19 2011
parent Andrew Wiley <wiley.andrew.j gmail.com> writes:
On Wed, Oct 19, 2011 at 9:38 PM, Robert Jacques <sandford jhu.edu> wrote:

 On Wed, 19 Oct 2011 20:51:25 -0400, Michel Fortin <
 michel.fortin michelf.com> wrote:

 On 2011-10-19 21:53:12 +0000, Andrew Wiley <wiley.andrew.j gmail.com>
 said:
[snip] Also, I how shared works is being misunderstood. Making a class synchronized should limit all member functions and field to being shared or immutable, but it doesn't place any limits on the argument to the member functions. So in the below: synchronized class Thing2 { void doSomeWork(int i) {} void doSomeOtherWork(Thing2 t) {} void work() {} } i is of type int, not shared(int).
Ah, I see, so I can pass unshared data or references to unshared data into member functions, and the type system keeps me from saving them to member fields because member fields will be flagged as shared, but the arguments need not be. That makes a lot more sense. Sorry for the noise.
Oct 19 2011