ref return function using foreach ref result segfaults. Compiler bug?
Hard to describe this problem, see code and read comments below.
class A
{
private int _v;
this( int a_v )
{
_v = a_v;
}
@property size_t length()
{
return 1;
}
int opApply( int delegate( ref int a_v ) a_dg )
{
int result = 0;
for ( ulong i = 0; i < length; ++i )
{
result = a_dg( this.opIndex( i ) );
if ( result ) break;
}
return result;
}
ref int opIndex( size_t a_iPos )
{
return _v;
}
}
class B : A
{
this( int a_v )
{
super(a_v);
}
ref int find( int a_What )
{
foreach( val; super )
{
if ( val == a_What )
return val;
}
throw new Exception("value not found");
}
}
main()
{
auto v_B = new B(500);
writefln("Search = %d", v_B.find(500) );
return 0;
}
When the return value of find() is ref, it segfaults or returns
garbage. If the return value is a copy it works OK.
The only oddity I can see is that 'val' goes out of scope, but
it's a ref return value (pointer) to _v (right?), so it should
work anyway.
This looks like a bug in the compiler to me.
What do you guys think?
--rt
Re: Compilable Recursive Data Structure ( was: Recursive data structure using template won't compile)
On 13/11/12 06:51, Rob T wrote: On Monday, 12 November 2012 at 14:28:53 UTC, Andrej Mitrovic wrote: On 11/12/12, Andrej Mitrovic wrote: On 11/12/12, Don Clugston wrote: Yeah. Though note that 1000 bug reports are from bearophile. Actually only around 300 remain open: http://d.puremagic.com/issues/buglist.cgi?query_format=advanced&emailreporter2=1&emailtype2=substring&order=Importance&bug_status=UNCONFIRMED&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED&bug_status=VERIFIED&email2=bearophile Oh wait, that's only for DMD. It's 559 in total: http://d.puremagic.com/issues/buglist.cgi?query_format=advanced&emailreporter2=1&emailtype2=substring&order=Importance&bug_status=UNCONFIRMED&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED&bug_status=VERIFIED&email2=bearophile Issue 8990 and 6969, both related to this thread and unresolved, are not in the list, so I suspect there's a lot more missing too. PS: I could not figure out how to make a useful report using that bug report tool either. --rt I recommend deskzilla lite. D is on its list of supported open-source projects. It maintains a local copy of the entire bugzilla database, so you're not restricted to the slow and horrible html interface.
Re: ref return function using foreach ref result segfaults. Compiler bug?
On Tuesday, 13 November 2012 at 08:50:16 UTC, Rob T wrote:
Hard to describe this problem, see code and read comments below.
class A
{
private int _v;
this( int a_v )
{
_v = a_v;
}
@property size_t length()
{
return 1;
}
int opApply( int delegate( ref int a_v ) a_dg )
{
int result = 0;
for ( ulong i = 0; i < length; ++i )
{
result = a_dg( this.opIndex( i ) );
if ( result ) break;
}
return result;
}
ref int opIndex( size_t a_iPos )
{
return _v;
}
}
class B : A
{
this( int a_v )
{
super(a_v);
}
ref int find( int a_What )
{
foreach( val; super )
{
if ( val == a_What )
return val;
}
throw new Exception("value not found");
}
}
main()
{
auto v_B = new B(500);
writefln("Search = %d", v_B.find(500) );
return 0;
}
When the return value of find() is ref, it segfaults or returns
garbage. If the return value is a copy it works OK.
The only oddity I can see is that 'val' goes out of scope, but
it's a ref return value (pointer) to _v (right?), so it should
work anyway.
This looks like a bug in the compiler to me.
What do you guys think?
This issue looks like bug8093.
http://d.puremagic.com/issues/show_bug.cgi?id=8093
And the code works correctly in git head (dmd2.061alpha).
Therefore, I think that the bug is fixed very recently.
Kenji Hara
Re: Compilable Recursive Data Structure ( was: Recursive data structure using template won't compile)
On 11/13/12, Don Clugston wrote: > I recommend deskzilla lite. D is on its list of supported open-source > projects. It maintains a local copy of the entire bugzilla database, so > you're not restricted to the slow and horrible html interface. Wow, I had no idea they had this. I've added a note about it here: http://prowiki.org/wiki4d/wiki.cgi?D__Tutorial/BugReports Thanks!
Re: Inferring function argument types from other argument types
I believe this question was asked before, but here is the
solution again.
struct Foo(_T1, _T2)
{
alias _T1 T1;
alias _T2 T2;
T1 a;
T2 b;
}
FooT.T1 func(FooT, T)(FooT foo, T x)
if (is(FooT.T1) && is(T : FooT.T1))
{
return x * foo.a;
}
void main() {
auto foo = Foo!(size_t, string)(8, "bobcat");
int value = 3;
assert(foo.func(value) == 24);
assert(is(typeof(foo.func(value)) == size_t));
}
- Vijay
On Monday, 12 November 2012 at 12:42:31 UTC, Joseph Rushton
Wakeling wrote:
Suppose that I've got a struct which internally defines a
number of types:
struct Foo(_T1, _T2)
{
alias _T1 T1;
alias _T2 T2;
T1 a;
T2 b;
}
... and now I want to define a function which takes as input an
instance of one of these structs, and a variable of type T1.
I tried the following:
T func(FooT, T = FooT.T1)(FooT foo, T x)
{
return x * foo.a;
}
but found that the type of whatever I was passing would
override the default: e.g. if I called func(fooInstance, 1)
then the second argument would be interpreted as an int even
though Foo.T1 is size_t.
I also tried,
T func(FooT, T : FooT.T1)(FooT foo, T x)
{
return x * foo.a;
}
but this generates a different error: "no property 'T1' for
type 'FooT'". I tried replacing FooT with alias FooT to no
avail.
Obviously I could get round the problems of the first example
with something based around CommonType etc. but I'm just
wondering if it's possible to make an argument dependent on
another template parameter in the way I'm looking for here.
Re: Regarding ranges
On Tuesday, 13 November 2012 at 04:50:54 UTC, bearophile wrote: Do you know if it's possible to write similar code nicely & efficiently with ranges? Bearophile, do you know of a good reference to learn about ranges in D? Due to the language name, Googling for specific topics in D is a bit difficult and though I've heard the topic come up, I never found a way to really understand it. - Vijay
Re: Compilable Recursive Data Structure ( was: Recursive data structure using template won't compile)
On Tue, 13 Nov 2012 11:56:57 +0100 Don Clugston wrote: > > I recommend deskzilla lite. D is on its list of supported open-source > projects. It maintains a local copy of the entire bugzilla database, > so you're not restricted to the slow and horrible html interface. > Awesome! I wish GitHub had something like that. (Hell, I wish every web "app" had something like that.) Although, $189 for anything with > 2k issues or anything non-OSS? Geez that's expensive, what are they, Adobe?
Shared value types can not be destructed
Apperently this is by design: http://d.puremagic.com/issues/show_bug.cgi?id=8295 To clarify: It is not possible to define a destructor that will be called on the destruction of a shared struct. In a different thread Walter commeted this bug with: "If you include an object designed to work only in a single thread (non-shared), make it shared, and then destruct it when other threads may be pointing to it ... What should happen?" I did try to think of a case where his scenario would actually break something, but couldn't find one. If someone has more knowdelge about this situation some clarification would be great. Maybe even a small code sample that illustrates the problem. Kind Regards Benjamin Thaut
Re: Regarding ranges
On Tuesday, 13 November 2012 at 16:18:43 UTC, Vijay Nayar wrote: On Tuesday, 13 November 2012 at 04:50:54 UTC, bearophile wrote: Do you know if it's possible to write similar code nicely & efficiently with ranges? Bearophile, do you know of a good reference to learn about ranges in D? Due to the language name, Googling for specific topics in D is a bit difficult and though I've heard the topic come up, I never found a way to really understand it. - Vijay http://ddili.org/ders/d.en/ranges.html
Re: Compilable Recursive Data Structure ( was: Recursive data structure using template won't compile)
On Tue, 13 Nov 2012 11:19:36 -0500 Nick Sabalausky wrote: > On Tue, 13 Nov 2012 11:56:57 +0100 > Don Clugston wrote: > > > > I recommend deskzilla lite. D is on its list of supported > > open-source projects. It maintains a local copy of the entire > > bugzilla database, so you're not restricted to the slow and > > horrible html interface. > > > > Awesome! I wish GitHub had something like that. (Hell, I wish every > web "app" had something like that.) > > Although, $189 for anything with > 2k issues or anything non-OSS? Geez > that's expensive, what are they, Adobe? > Oh, upon using it, it looks like the free version is for OSS projects with any number of issues *and* non-OSS projects with <= 2k issues. Guess I misread their download page.
Re: Inferring function argument types from other argument types
On 11/13/2012 05:05 PM, Vijay Nayar wrote:
I believe this question was asked before, but here is the solution again.
The actual reality of what I'm trying to do is slightly more complex: it's more
like
struct Foo(_T1, _T2)
{
alias _T1 T1;
alias _T2 T2;
// etc.
}
FooT.T1 func(FooT, T)(FooT foo, T x)
{
return func2(x);
}
... where func2() is also a templated function, and it's important that it take
a type of FooT.T1 and not T.
So, I can't see what the solution is apart from casting x to FooT.T or
explicitly indicating FooT.T1 as a template parameter for func2.
Re: ref return function using foreach ref result segfaults. Compiler bug?
On Tuesday, 13 November 2012 at 12:31:26 UTC, Kenji Hara wrote: This issue looks like bug8093. http://d.puremagic.com/issues/show_bug.cgi?id=8093 And the code works correctly in git head (dmd2.061alpha). Therefore, I think that the bug is fixed very recently. Kenji Hara Thanks for the response Kenji! I'll implement a work-a-round for now and won't bother filing a bug report. --rt
Re: Inferring function argument types from other argument types
Ok, I get it.
My understanding is that you have basically two options.
* If you want the function to be called with any implicitly
castable type, then you must cast it in the function.
* If you only want the function to accept the one type,
then use is(T == FooT.T1) and you must cast the input
the function.
However, all this feels like an over complication and round-about
way of
doing things. If you have exact types that must be specified in
the
struct, you probably don't want an external template at all.
Simply place
the function in the struct.
struct Foot(_T1, _T2) {
...
_T1 func(_T1 x) {
return func2(x);
}
}
- Vijay
On Tuesday, 13 November 2012 at 16:39:55 UTC, Joseph Rushton
Wakeling wrote:
On 11/13/2012 05:05 PM, Vijay Nayar wrote:
I believe this question was asked before, but here is the
solution again.
The actual reality of what I'm trying to do is slightly more
complex: it's more like
struct Foo(_T1, _T2)
{
alias _T1 T1;
alias _T2 T2;
// etc.
}
FooT.T1 func(FooT, T)(FooT foo, T x)
{
return func2(x);
}
... where func2() is also a templated function, and it's
important that it take a type of FooT.T1 and not T.
So, I can't see what the solution is apart from casting x to
FooT.T or explicitly indicating FooT.T1 as a template parameter
for func2.
Re: Inferring function argument types from other argument types
On 11/12/2012 06:55 AM, Joseph Rushton Wakeling wrote:
> I'm just curious
> if there is any other way to meaningfully determine the type of one
> function argument based on the type of another. Hence the playing with
> different template argument formulations.
>
> It just feels really difficult to believe that the generic programming
> in D doesn't extend to allowing you to infer one argument type based on
> another.
I am pretty sure that should work according to spec under "Argument
Deduction":
http://dlang.org/template.html
"
2. If the type specialization is dependent on a type parameter, the type
of that parameter is set to be the corresponding part of the type argument.
"
struct Foo(_T1, _T2)
{
alias _T1 T1;
T1 t1;
}
void foo(T)(T foo, T.T1 x)// <-- Note T.T1
{
return func2(x);
}
void main()
{
auto f = Foo!(size_t, string)();
foo(f, 42);
}
Although, there is the issue of compiler's not being sure whether the T1
in T.T1 is a typename or a member variable. (That's why the 'typename'
keyword is used in such contexts in C++.)
Still, please create a bug report: :)
http://d.puremagic.com/issues/
Ali
Re: Regarding ranges
This is very well written. Thanks for the link! - Vijay On Tuesday, 13 November 2012 at 16:29:41 UTC, Namespace wrote: On Tuesday, 13 November 2012 at 16:18:43 UTC, Vijay Nayar wrote: On Tuesday, 13 November 2012 at 04:50:54 UTC, bearophile wrote: Do you know if it's possible to write similar code nicely & efficiently with ranges? Bearophile, do you know of a good reference to learn about ranges in D? Due to the language name, Googling for specific topics in D is a bit difficult and though I've heard the topic come up, I never found a way to really understand it. - Vijay http://ddili.org/ders/d.en/ranges.html
Extent of tail call optimization in D?
In Dr Alexandrescu's "The D Programming Language", on page 12, it
is noted that the D compiler will rewrite tail calls within a
procedure as loops. Does the compiler rewrite tail calls between
procedures as jumps?
For example, in pseudo-D:
void foo(K)(K cont, stuff..) {
// Do things with stuff.
cont(things);
}
Is the call `cont(things)' optimized?
Re: Is there a way to initialize a non-assigned structure declaration (or is it a definition)?
This is merely a syntactic difference in how structs are handled.
In D, structs are more akin to low level types like int and have
most of the same symantics.
So
SnonParameterized cnp(5, 3.303);
makes about as much sense as
int cnp(3);
You have two syntax choices to pick from in the D version of the
code you posted:
struct SnonParameterized
{
int t;
float u;
};
void main() {
auto snon1 = SnonParameterized(3, 4.5); // Option 1
SnonParameterized snon2 = {3, 4.5}; // Option 2
}
- Vijay
On Saturday, 10 November 2012 at 17:41:00 UTC, Too Embarrassed To
Say wrote:
I appreciate all the helpful replies, but I've simplified
things to what I belive is the core issue. In C++ (at the risk
of becoming a heretic) the language allows me to do the
following:
struct SnonParameterized
{
public:
int t;
float u;
SnonParameterized(int tparam, float uparam);
};
SnonParameterized::SnonParameterized(int tparam, float uparam)
{
t = tparam;
u = uparam;
}
SnonParameterized snp(5, 3.303); // this compiles with Visual
C++ 2010
===
Now with D, I try (what I think is identical semantics) the
following:
struct SnonParameterized
{
int t;
float u;
this(int t, float u)
{
this.t = t
this.u = u;
}
}
SnonParameterized cnp(5, 3.303); // fails compile with Error:
found 'cnp' when expecting ';' following statement
auto hi = SnonParameterized(5, 3.303); // compiles of course.
I'm just trying to understand why D disallows the
non-assignment syntax. Probably for a very good (and obvious)
reason.
Re: Is Override Still Mandatory?
On Monday, 12 November 2012 at 03:08:35 UTC, Jonathan M Davis wrote: On Monday, November 12, 2012 03:58:17 Vijay Nayar wrote: I was under the impression that the attribute "override" was mandatory when replacing a virtual function in a base class. However, code that leaves this out has no errors using dmd v2.060. It's mandatory if you compile with -w. Eventually, it will be mandatory all the time. It's just that it's being phased in via -w rather than requiring it immediately. That way, people have time to fix their code to use it before it's required. Unfortunately though, a lot of people fail to compile with either -w or -wi, so they're happily writing _new_ code which will break once override is always required, which means that phasing like this doesn't necessarily actually fix the problem. - Jonathan M Davis I was about to ask this same question today! Thanks for the answer. --rt
Re: Is Override Still Mandatory?
On Monday, 12 November 2012 at 03:29:09 UTC, bearophile wrote: I agree. Generally I think warnings should be active on default and disabled on request with a switch :-) Bye, bearophile Agreed! Fortunately I have not written all that much broken code yet. I recall a lot of discussion about how "depreciated" should work, and it's clear what we have now is next to useless, so I hope the suggestion that was made to improve depreciated with warnings and so forth, is implemented. --rt
Re: Is Override Still Mandatory?
Rob T: so I hope the suggestion that was made to improve depreciated with warnings and so forth, is implemented. "Deprecated features as warnings" is in the D front since about 7 hours: https://github.com/D-Programming-Language/dmd/commit/5881617a34adc172b830314c17da21d5c834ffd0 Bye, bearophile
Re: Is Override Still Mandatory?
On Wednesday, 14 November 2012 at 02:33:47 UTC, bearophile wrote: Rob T: so I hope the suggestion that was made to improve depreciated with warnings and so forth, is implemented. "Deprecated features as warnings" is in the D front since about 7 hours: https://github.com/D-Programming-Language/dmd/commit/5881617a34adc172b830314c17da21d5c834ffd0 Bye, bearophile That's good news. If I were to use the latest pre-release version of dmd that was relatively safe, how will I find it, or is it OK to use the master branch? --rt
Re: Is Override Still Mandatory?
On Wednesday, 14 November 2012 at 02:33:47 UTC, bearophile wrote: Rob T: so I hope the suggestion that was made to improve depreciated with warnings and so forth, is implemented. "Deprecated features as warnings" is in the D front since about 7 hours: https://github.com/D-Programming-Language/dmd/commit/5881617a34adc172b830314c17da21d5c834ffd0 Bye, bearophile That's good news. If I were to use the latest pre-release version of dmd that was relatively safe, how will I find it, or is it OK to use the master branch? --rt
