http://ftp.digitalmars.com/dmd2beta.zip

Please take a look at the message I left on https://github.com/D-Programming-Language/dmd/commit/f330b950171adb67d2b648885c528b2fd5909130#comments

I think it's a regression.

On Mon, Feb 13, 2012 at 9:01 PM, Walter Bright <walter at digitalmars.com> wrote:
> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 13 feb 2012, at 11:01, Walter Bright wrote:

> http://ftp.digitalmars.com/dmd2beta.zip


My serialization library Orange doesn't work with this beta. I get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.

This is the part of the GDB backtrace that repeats itself.

#3991 0x0000000100015f6f in D6orange4util3Use79__T7restoreTvTS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4NodeZ7restoreFKS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4NodeZS6orange4util3Use86__T13RestoreStructTvTS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4NodeZ13RestoreStruct10__lambda26MFDFZvKS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4NodeZv ()
#3992 0x0000000100015ea4 in D6orange4util3Use86__T13RestoreStructTvTS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4NodeZ13RestoreStruct4opInMFDFZvZv ()
#3993 0x0000000100005a36 in D6orange13serialization8archives10XmlArchive18__T10XmlArchiveTaZ10XmlArchive15unarchiveObjectMFAyaJmJC6ObjectDFZvZv ()
#3994 0x0000000100009ec6 in D6orange13serialization10Serializer10Serializer38__T17deserializeObjectTC4main3FooTAyaZ17deserializeObjectMFAyaZC4main3Foo ()
#3995 0x0000000100009de7 in D6orange13serialization10Serializer10Serializer40__T19deserializeInternalTC4main3FooTAyaZ19deserializeInternalMFAyaZC4main3Foo ()
#3996 0x000000010000a1ac in D6orange13serialization10Serializer10Serializer46__T29objectStructDeserializeHelperTC4main3FooZ29objectStructDeserializeHelperMFKC4main3FooZv ()
#3997 0x000000010000a0eb in D6orange13serialization10Serializer10Serializer38__T17deserializeObjectTC4main3FooTAyaZ17deserializeObjectMFAyaZC4main3Foo10__lambda15MFZv10__lambda16MFZv ()
#3998 0x0000000100009d68 in D6orange13serialization10Serializer10Serializer30__T13triggerEventsTC4main3FooZ13triggerEventsMFC4main3FooDFZvZv ()
#3999 0x0000000100009f18 in D6orange13serialization10Serializer10Serializer38__T17deserializeObjectTC4main3FooTAyaZ17deserializeObjectMFAyaZC4main3Foo10__lambda15MFZv ()

-- 
/Jacob Carlborg

On 2/13/12 9:03 PM, Jacob Carlborg wrote:
> My serialization library Orange doesn't work with this beta. I get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.

Could you put together a test case? __lambda appears in the mangled name of, well, lambdas ? or do you mean that you aren't using any function literals in your code?

David

On 13 feb 2012, at 21:08, David Nadlinger wrote:

> On 2/13/12 9:03 PM, Jacob Carlborg wrote:
>> My serialization library Orange doesn't work with this beta. I get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.
> 
> Could you put together a test case?

That's what I'm working on.

> __lambda appears in the mangled name of, well, lambdas ? or do you mean that you aren't using any function literals in your code?
> 
> David


I'm using delegates, but not the new lambda syntax, () => foo;

-- 
/Jacob Carlborg

On 2/13/2012 12:03 PM, Jacob Carlborg wrote:
> My serialization library Orange doesn't work with this beta. I get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.
>
> This is the part of the GDB backtrace that repeats itself.


I'm sorry, but a back trace from your code is not useful to figuring out what went wrong. I don't have your code, know what you're doing, when did this problem appear, anything.

On 2/13/2012 3:05 AM, Daniel Murphy wrote:
> Please take a look at the message I left on https://github.com/D-Programming-Language/dmd/commit/f330b950171adb67d2b648885c528b2fd5909130#comments
>
> I think it's a regression.
>

I have a hard time mentally connecting which name is used on the mailing list with which pseudonym on github. I know if I click on yebblies I get Daniel Murphy, but it would be more convenient if you used DanielMurphy as the github handle. This applies to everyone.

And btw, using your own name for the commits will increase your professional ranking in Google searches, as Google will be able to find your contributions easily :-).

On 13 feb 2012, at 21:16, Walter Bright wrote:

> 
> 
> On 2/13/2012 12:03 PM, Jacob Carlborg wrote:
>> My serialization library Orange doesn't work with this beta. I get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.
>> 
>> This is the part of the GDB backtrace that repeats itself.
> 
> 
> I'm sorry, but a back trace from your code is not useful to figuring out what went wrong. I don't have your code, know what you're doing, when did this problem appear, anything.


Yeah, I know, sorry. I just posted so you guys know while I'm working on a test case. Hmm, when I'm looking at the backtrace it might be another problem. It might be a problem with array appending or the GC, in this method:

https://github.com/jacob-carlborg/orange/blob/master/orange/xml/PhobosXml.d#L764

The above method is the last step in the backtrace that is part of my code.

This is the start of the backtrace:

#0  0x0000000100042ed0 in D2gc3gcx2GC11queryNoSyncMFPvZS2gc3gcx7BlkInfo ()
#1  0x0000000100042e35 in D2gc3gcx2GC5queryMFPvZS2gc3gcx7BlkInfo ()
#2  0x00000001000414b0 in gc_query ()
#3  0x000000010004ae0d in _d_arrayappendcTX ()
#4  0x000000010000b7f6 in D6orange3xml9PhobosXml7Element10attributesMFZAC6orange3xml9PhobosXml9Attribute17__foreachbody1522MFKAyaKAyaZi ()
#5  0x0000000100046eae in _aaApply2 ()
#6  0x000000010000b76c in D6orange3xml9PhobosXml7Element10attributesMFZAC6orange3xml9PhobosXml9Attribute ()
#7  0x0000000100001c67 in D6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4Node10attributesMFZS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument12VisitorProxy ()
#8  0x0000000100002042 in D6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument10QueryProxy9attributeMFDFS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4NodeZbZS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument10QueryProxy ()
#9  0x0000000100006cab in D6orange13serialization8archives10XmlArchive18__T10XmlArchiveTaZ10XmlArchive10getElementMFAyaAyaAyabZS6orange3xml11XmlDocument19__T11XmlDocumentTaZ11XmlDocument4Node ()

-- 
/Jacob Carlborg

On 13 feb 2012, at 21:31, Jacob Carlborg wrote:

> 
> On 13 feb 2012, at 21:16, Walter Bright wrote:
> 
>> 
>> 
>> On 2/13/2012 12:03 PM, Jacob Carlborg wrote:
>>> My serialization library Orange doesn't work with this beta. I get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.
>>> 
>>> This is the part of the GDB backtrace that repeats itself.
>> 
>> 
>> I'm sorry, but a back trace from your code is not useful to figuring out what went wrong. I don't have your code, know what you're doing, when did this problem appear, anything.
> 
> 
> Yeah, I know, sorry. I just posted so you guys know while I'm working on a test case. Hmm, when I'm looking at the backtrace it might be another problem. It might be a problem with array appending or the GC, in this method:
> 
> https://github.com/jacob-carlborg/orange/blob/master/orange/xml/PhobosXml.d#L764


I'm seeing some general problem with array appending. Removing the append in the above method and the segfault moves to an other part of the code, which uses array appending.

-- 
/Jacob Carlborg

This can be any number of things.? It seems like a corruption issue.


Just for your info, however, the code you quoted below is wasteful -- Starting with 2.041 or around there, the code you give will allocate a new array, then throw it away on the first append.

You are better off with this:

Attribute[] attrs;
attrs.reserve(tag.attr.length);

No, I don't think this will solve your problem.? But it should avoid the throw-away initial array allocation.


-Steve



----- Original Message -----
> From: Jacob Carlborg <doob at me.com>
> To: Discuss the dmd beta releases for D <dmd-beta at puremagic.com>
> Cc:
> Sent: Monday, February 13, 2012 3:47 PM
> Subject: Re: [dmd-beta] D2 2.058 beta 4
> 
> 
> On 13 feb 2012, at 21:31, Jacob Carlborg wrote:
> 
>> 
>>  On 13 feb 2012, at 21:16, Walter Bright wrote:
>> 
>>> 
>>> 
>>>  On 2/13/2012 12:03 PM, Jacob Carlborg wrote:
>>>>  My serialization library Orange doesn't work with this beta. I
> get a segmentation fault. The backtrace from GDB seems to indicate some sort of infinite loop or recursion. I see some traces of "__lambda" in the backtrace but that is not used anywhere in my code.
>>>> 
>>>>  This is the part of the GDB backtrace that repeats itself.
>>> 
>>> 
>>>  I'm sorry, but a back trace from your code is not useful to
> figuring out what went wrong. I don't have your code, know what you're doing, when did this problem appear, anything.
>> 
>> 
>>  Yeah, I know, sorry. I just posted so you guys know while I'm working
> on a test case. Hmm, when I'm looking at the backtrace it might be another problem. It might be a problem with array appending or the GC, in this method:
>> 
>> 
> https://github.com/jacob-carlborg/orange/blob/master/orange/xml/PhobosXml.d#L764
> 
> 
> I'm seeing some general problem with array appending. Removing the append in the above method and the segfault moves to an other part of the code, which uses array appending.
> 
> -- 
> /Jacob Carlborg
> 
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>