www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Got compiler crash in Mangler::mangleFuncType

reply Uranuz <neuranuz gmail.com> writes:
After changes in my project tried to compile, but compiler 
failed. Running dmd under GDB gives the following:
Program received signal SIGSEGV, Segmentation fault.
0x00000000005807a0 in Mangler::mangleFuncType(TypeFunction*, 
TypeFunction*, unsigned char, Type*) ()

Could someone give me adivice how to investigate and maybe fix 
this problem?
Mar 24 2017
next sibling parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 After changes in my project tried to compile, but compiler 
 failed. Running dmd under GDB gives the following:
 Program received signal SIGSEGV, Segmentation fault.
 0x00000000005807a0 in Mangler::mangleFuncType(TypeFunction*, 
 TypeFunction*, unsigned char, Type*) ()

 Could someone give me adivice how to investigate and maybe fix 
 this problem?


Tried to build DMD from source using manual:
https://wiki.dlang.org/Building_under_Posix

Running compiled DMD to build my project gives the following 
error:
core.exception.AssertError ddmd/func.d(1641): Assertion failure
----------------
??:? _d_assert [0x71c730]
??:? void ddmd.func.__assert(int) [0x5b52de]
??:? _ZN15FuncDeclaration9semantic3EP5Scope [0x5ab904]
??:? _ZN20AggregateDeclaration9semantic3EP5Scope [0x500a1f]
??:? _ZN16TemplateInstance9semantic3EP5Scope [0x5705fd]
??:? _ZN6Module9semantic3EP5Scope [0x552b19]
??:? int ddmd.mars.tryMain(ulong, const(char)**) [0x5da406]
??:? _Dmain [0x5db2c6]
??:? 
_D2rt6dmain211_d_run_mainUiPPaPUAAaZiZ6runAllMFZ9__lambda1MFZv 
[0x71e5d7]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int 
function(char[][])*).tryExec(scope void delegate()) [0x71e503]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int 
function(char[][])*).runAll() [0x71e57c]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int 
function(char[][])*).tryExec(scope void delegate()) [0x71e503]
??:? _d_run_main [0x71e467]
??:? main [0x5dba67]
??:? __libc_start_main [0x2c5a82f]
Mar 25 2017
parent Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Saturday, 25 March 2017 at 07:16:34 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 [...]


 Tried to build DMD from source using manual:
 https://wiki.dlang.org/Building_under_Posix

 [...]


try using Dustmite (https://github.com/CyberShadow/DustMite/) to 
reduce your project to the minimal amount that still reproduces 
the crash.
Mar 25 2017
prev sibling next sibling parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 After changes in my project tried to compile, but compiler 
 failed. Running dmd under GDB gives the following:
 Program received signal SIGSEGV, Segmentation fault.
 0x00000000005807a0 in Mangler::mangleFuncType(TypeFunction*, 
 TypeFunction*, unsigned char, Type*) ()

 Could someone give me adivice how to investigate and maybe fix 
 this problem?


Switched to v2.073.2. Used this built version of compiler and got 
some additional messages from compiler:
src/mkk_site/view_service/access_control.d(25): Deprecation: 
mkk_site.view_service.utils.HTTPContext is not visible from 
module access_control
../ivy/src/ivy/interpreter.d-mixin-670(670): Error: undefined 
identifier '__cmp' in module 'object'
../ivy/src/ivy/interpreter.d-mixin-670(670): Error: undefined 
identifier '__cmp' in module 'object'
../ivy/src/ivy/interpreter.d-mixin-670(670): Error: undefined 
identifier '__cmp' in module 'object'
../ivy/src/ivy/interpreter.d-mixin-670(670): Error: undefined 
identifier '__cmp' in module 'object'
core.exception.AssertError ddmd/func.d(1641): Assertion failure
----------------
??:? _d_assert [0x71c730]
??:? void ddmd.func.__assert(int) [0x5b52de]
??:? _ZN15FuncDeclaration9semantic3EP5Scope [0x5ab904]
??:? _ZN20AggregateDeclaration9semantic3EP5Scope [0x500a1f]
??:? _ZN16TemplateInstance9semantic3EP5Scope [0x5705fd]
??:? _ZN6Module9semantic3EP5Scope [0x552b19]
??:? int ddmd.mars.tryMain(ulong, const(char)**) [0x5da406]
??:? _Dmain [0x5db2c6]
??:? 
_D2rt6dmain211_d_run_mainUiPPaPUAAaZiZ6runAllMFZ9__lambda1MFZv 
[0x71e5d7]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int 
function(char[][])*).tryExec(scope void delegate()) [0x71e503]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int 
function(char[][])*).runAll() [0x71e57c]
??:? void rt.dmain2._d_run_main(int, char**, extern (C) int 
function(char[][])*).tryExec(scope void delegate()) [0x71e503]
??:? _d_run_main [0x71e467]
??:? main [0x5dba67]
??:? __libc_start_main [0xcc85082f]
Mar 25 2017
parent Jacob Carlborg <doob me.com> writes:
On 2017-03-25 08:25, Uranuz wrote:


 ../ivy/src/ivy/interpreter.d-mixin-670(670): Error: undefined identifier
 '__cmp' in module 'object'


I think this error is due to a mismatch of DMD and druntime. The 
compiler lowers some kind of comparison to a call to __cmp which is 
implemented in the "object" module in druntime. This is a quite recent 
change.

-- 
/Jacob Carlborg
Mar 25 2017
prev sibling parent reply Stefan Koch <uplink.coder googlemail.com> writes:
On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 After changes in my project tried to compile, but compiler 
 failed. Running dmd under GDB gives the following:
 Program received signal SIGSEGV, Segmentation fault.
 0x00000000005807a0 in Mangler::mangleFuncType(TypeFunction*, 
 TypeFunction*, unsigned char, Type*) ()

 Could someone give me adivice how to investigate and maybe fix 
 this problem?


Switch to git ~master and see if it still occurs.
Also if you can share the source I am happy to debug this.
Mar 25 2017
next sibling parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 After changes in my project tried to compile, but compiler 
 failed. Running dmd under GDB gives the following:
 Program received signal SIGSEGV, Segmentation fault.
 0x00000000005807a0 in Mangler::mangleFuncType(TypeFunction*, 
 TypeFunction*, unsigned char, Type*) ()

 Could someone give me adivice how to investigate and maybe fix 
 this problem?


 Switch to git ~master and see if it still occurs.
 Also if you can share the source I am happy to debug this.


I just put debug messages on every `assert` and figured out it 
fails on lines:
                     bool implicit0 = (f.next.ty == Tvoid && 
isMain());
                     Type tret = implicit0 ? Type.tint32 : f.next;
                     assert(tret.ty != Tvoid);
                     if (vresult || returnLabel)
                         buildResultVar(scout ? scout : sc2, tret);
I really don't understand what it is... Now I cannot share 
codebase easily. Seems that it fails on my programme `main` 
function or near that, because compiler stack not so deep. There 
is some check for isMain AFAIK... What is this about?
Mar 25 2017
next sibling parent reply Stefan Koch <uplink.coder googlemail.com> writes:
On Saturday, 25 March 2017 at 08:13:08 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 [...]


 I just put debug messages on every `assert` and figured out it 
 fails on lines:
                     bool implicit0 = (f.next.ty == Tvoid && 
 isMain());
                     Type tret = implicit0 ? Type.tint32 : 
 f.next;
                     assert(tret.ty != Tvoid);
                     if (vresult || returnLabel)
                         buildResultVar(scout ? scout : sc2, 
 tret);
 I really don't understand what it is... Now I cannot share 
 codebase easily. Seems that it fails on my programme `main` 
 function or near that, because compiler stack not so deep. 
 There is some check for isMain AFAIK... What is this about?


are you creating a function called main ?
that would be about the only time this would be triggered.
Mar 25 2017
parent Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 08:24:04 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 08:13:08 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 [...]


 I just put debug messages on every `assert` and figured out it 
 fails on lines:
                     bool implicit0 = (f.next.ty == Tvoid && 
 isMain());
                     Type tret = implicit0 ? Type.tint32 : 
 f.next;
                     assert(tret.ty != Tvoid);
                     if (vresult || returnLabel)
                         buildResultVar(scout ? scout : sc2, 
 tret);
 I really don't understand what it is... Now I cannot share 
 codebase easily. Seems that it fails on my programme `main` 
 function or near that, because compiler stack not so deep. 
 There is some check for isMain AFAIK... What is this about?


 are you creating a function called main ?
 that would be about the only time this would be triggered.


Seems that compiler thinks that `struct` constructor returns some 
value, but declared type of return value for struct ctor is 
Tvoid, so it fails on assert. Printing returns.toChars() after 
condition `if (returns)`  gives `[return this;`. Maybe `returns` 
should be null there or not? I'm very confident that it fails on 
constructor. Tried to reduce example manually (I'm not very 
confident about Dustmite). But it doesn't reproduce error:

import std.stdio;

struct DirValueAttr(bool isForCompiler = false)
{
	string name;
	string typeName;

	static if( isForCompiler )
	{
		import ivy.node: IExpression;

		IExpression defaultValueExpr;
		this( string name, string typeName, IExpression defValue )
		{
			this.name = name;
			this.typeName = typeName;
			this.defaultValueExpr = defValue;
		}
	}
	else
	{
		this( string name, string typeName )
		{
			this.name = name;
			this.typeName = typeName;
		}
	}

	DirValueAttr!(false) toInterpreterValue() {
		return DirValueAttr!(false)(name, typeName);
	}
}

// We pass callable attributes by blocks. Every attribute block 
has size an type
// Size and type stored in single integer argument in stack 
preceding the block
// Size is major binary part of this integer denoted by bit 
offset:
enum size_t _stackBlockHeaderSizeOffset = 4;
// To have some validity check bit between size and block type 
must always be zero
// The following mask is used to check for validity:
enum size_t _stackBlockHeaderCheckMask = 0b1000;
// And there is mask to extract type of block
enum size_t _stackBlockHeaderTypeMask = 0b111;


enum DirAttrKind { NamedAttr, ExprAttr, IdentAttr, KwdAttr, 
BodyAttr };

static this()
{
	static assert( DirAttrKind.max <= _stackBlockHeaderTypeMask, 
`DirAttrKind set of values exeeded of defined limit` );
}

struct DirAttrsBlock(bool isForCompiler = false)
{
	import std.typecons: Tuple, tuple;
	alias TValueAttr = DirValueAttr!(isForCompiler);

	static if( isForCompiler ) {
		import ivy.node: ICompoundStatement;
		alias TBodyTuple = Tuple!(ICompoundStatement, "ast", bool, 
"isNoscope");
	}
	else {
		alias TBodyTuple = Tuple!(bool, "isNoscope");
	}

	static struct Storage {
		union {
			TValueAttr[string] namedAttrs;
			TValueAttr[] exprAttrs;
			string[] names;
			string keyword;
			TBodyTuple bodyAttr;
		}
	}

	private DirAttrKind _kind;
	private Storage _storage;

	this( DirAttrKind attrKind, TValueAttr[string] attrs )
	{
		assert( attrKind == DirAttrKind.NamedAttr, `Expected NamedAttr 
kind for attr block` );

		_kind = attrKind;
		_storage.namedAttrs = attrs;
	}

	this( DirAttrKind attrKind, TValueAttr[] attrs )
	{
		assert( attrKind == DirAttrKind.ExprAttr, `Expected ExprAttr 
kind for attr block` );

		_kind = attrKind;
		_storage.exprAttrs = attrs;
	}

	this( DirAttrKind attrKind, string[] names )
	{
		assert( attrKind == DirAttrKind.IdentAttr, `Expected IdentAttr 
kind for attr block` );

		_kind = attrKind;
		_storage.names = names;
	}

	this( DirAttrKind attrKind, string kwd )
	{
		assert( attrKind == DirAttrKind.KwdAttr, `Expected Keyword kind 
for attr block` );

		_kind = attrKind;
		_storage.keyword = kwd;
	}

	this( DirAttrKind attrKind, TBodyTuple value )
	{
		assert( attrKind == DirAttrKind.BodyAttr, `Expected BodyAttr 
kind for attr block` );

		_kind = attrKind;
		_storage.bodyAttr = value;
	}

	this( DirAttrKind attrKind )
	{
		_kind = attrKind;
	}

	DirAttrKind kind()  property {
		return _kind;
	}

	void kind(DirAttrKind value)  property {
		_kind = value;
	}

	void namedAttrs( TValueAttr[string] attrs )  property {
		_storage.namedAttrs = attrs;
		_kind = DirAttrKind.NamedAttr;
	}

	TValueAttr[string] namedAttrs()  property {
		assert( _kind == DirAttrKind.NamedAttr, `Directive attrs block 
is not of NamedAttr kind` );
		return _storage.namedAttrs;
	}

	void exprAttrs( TValueAttr[] attrs )  property {
		_storage.exprAttrs = attrs;
		_kind = DirAttrKind.ExprAttr;
	}

	TValueAttr[] exprAttrs()  property {
		assert( _kind == DirAttrKind.ExprAttr, `Directive attrs block 
is not of ExprAttr kind` );
		return _storage.exprAttrs;
	}

	void names(string[] names)  property {
		_storage.names = names;
		_kind = DirAttrKind.IdentAttr;
	}

	string[] names()  property {
		assert( _kind == DirAttrKind.IdentAttr, `Directive attrs block 
is not of IdentAttr kind` );
		return _storage.names;
	}

	void keyword(string value)  property {
		_storage.keyword = value;
		_kind = DirAttrKind.KwdAttr;
	}

	string keyword()  property {
		assert( _kind == DirAttrKind.KwdAttr, `Directive attrs block is 
not of KwdAttr kind` );
		return _storage.keyword;
	}

	void bodyAttr(TBodyTuple value)  property {
		_storage.bodyAttr = value;
		_kind = DirAttrKind.BodyAttr;
	}

	TBodyTuple bodyAttr()  property {
		assert( _kind == DirAttrKind.BodyAttr, `Directive attrs block 
is not of BodyAttr kind` );
		return _storage.bodyAttr;
	}

	DirAttrsBlock!(false) toInterpreterBlock()
	{
		import std.algorithm: map;
		import std.array: array;

		final switch( _kind )
		{
			case DirAttrKind.NamedAttr: {
				DirValueAttr!(false)[string] attrs;
				foreach( key, ref currAttr; _storage.namedAttrs ) {
					attrs[key] = currAttr.toInterpreterValue();
				}
				return DirAttrsBlock!(false)( _kind, attrs );
			}
			case DirAttrKind.ExprAttr:
				return DirAttrsBlock!(false)( _kind, _storage.exprAttrs.map!( 
a => a.toInterpreterValue() ).array );
			case DirAttrKind.IdentAttr:
				return DirAttrsBlock!(false)( _kind, _storage.names );
			case DirAttrKind.KwdAttr:
				return DirAttrsBlock!(false)( _kind, _storage.keyword );
			case DirAttrKind.BodyAttr:
				return DirAttrsBlock!(false)( _kind, 
tuple!("isNoscope")(_storage.bodyAttr.isNoscope) );
		}
		assert( false, `This should never happen` );
	}

	string toString()
	{
		import std.conv: to;
		final switch( _kind ) with( DirAttrKind )
		{
			case NamedAttr:
			case ExprAttr:
			case IdentAttr:
			case KwdAttr:
			case BodyAttr:
				return `<` ~ _kind.to!string ~ ` attrs block>`;
		}
	}
}
void func() {
	DirAttrsBlock!(false)[] _attrBlocks = [
		DirAttrsBlock!false( DirAttrKind.NamedAttr, [
			"__result__": DirValueAttr!(false)( "__result__", "any" )
		]),
		DirAttrsBlock!false( DirAttrKind.BodyAttr )
	];

}

void main()
{
	func();

}
In full project it fails on `this( DirAttrKind attrKind, 
TValueAttr[string] attrs )` ctor.
Mar 25 2017
prev sibling next sibling parent Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Saturday, 25 March 2017 at 08:13:08 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 [...]


 Switch to git ~master and see if it still occurs.
 Also if you can share the source I am happy to debug this.


 I just put debug messages on every `assert` and figured out it 
 fails on lines:
                     bool implicit0 = (f.next.ty == Tvoid && 
 isMain());
                     Type tret = implicit0 ? Type.tint32 : 
 f.next;
                     assert(tret.ty != Tvoid);
                     if (vresult || returnLabel)
                         buildResultVar(scout ? scout : sc2, 
 tret);
 I really don't understand what it is... Now I cannot share 
 codebase easily.


dust mite has an "--obfuscate" option exactly for when you can't 
share the code.
see also the relevant section of the dust mite wiki
https://github.com/CyberShadow/DustMite/wiki/Detecting-a-segfault-in-dmd-itself


 Seems that it fails on my programme `main` function or near 
 that, because compiler stack not so deep. There is some check 
 for isMain AFAIK... What is this about?
Mar 25 2017
prev sibling parent Jacob Carlborg <doob me.com> writes:
On 2017-03-25 09:13, Uranuz wrote:


 I just put debug messages on every `assert` and figured out it fails on
 lines:
                     bool implicit0 = (f.next.ty == Tvoid && isMain());
                     Type tret = implicit0 ? Type.tint32 : f.next;
                     assert(tret.ty != Tvoid);
                     if (vresult || returnLabel)
                         buildResultVar(scout ? scout : sc2, tret);
 I really don't understand what it is... Now I cannot share codebase
 easily. Seems that it fails on my programme `main` function or near
 that, because compiler stack not so deep. There is some check for isMain
 AFAIK... What is this about?


To me it looks like there's some mismatch of the declared return type of 
"main" and what is actually returned. In D it's possible to declare 
"main" to return "void" and the compiler/runtime will automatically 
handle and infer the exit code of the program. Can you show just the 
"main" function?

-- 
/Jacob Carlborg
Mar 25 2017
prev sibling parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 06:56:40 UTC, Uranuz wrote:

 After changes in my project tried to compile, but compiler 
 failed. Running dmd under GDB gives the following:
 Program received signal SIGSEGV, Segmentation fault.
 0x00000000005807a0 in Mangler::mangleFuncType(TypeFunction*, 
 TypeFunction*, unsigned char, Type*) ()

 Could someone give me adivice how to investigate and maybe fix 
 this problem?


 Switch to git ~master and see if it still occurs.
 Also if you can share the source I am happy to debug this.


If you wish to help. You are welcome) I created branch in my 
repository `https://github.com/uranuz/ivy` called 
`v0.1/fail_on_compiler_error`. Seems that error depends on the 
order of source files that were passed to DMD. The way that error 
occurs is using RDMD utility, by running command(in project root 
folder):
rdmd "-Isrc/" -ofsimple_run_test test/simple_run_test.d

But if I run it using `dub` it compiles successfully(config is in 
dub.json):
dub run -c simple_run_test

Printing actual dmd command, used to compile project shows that 
dub puts `simple_run_test.d` at the end of source files list, but 
rdmd puts it at the beginning. And DMD fails in the first case 
without printing any adequate error message? Is there any ideas 
about it, besides using `dub` everywhere?)..
Mar 25 2017
parent reply Stefan Koch <uplink.coder googlemail.com> writes:
On Saturday, 25 March 2017 at 13:34:43 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 [...]


 If you wish to help. You are welcome) I created branch in my 
 repository `https://github.com/uranuz/ivy` called 
 `v0.1/fail_on_compiler_error`. Seems that error depends on the 
 order of source files that were passed to DMD. The way that 
 error occurs is using RDMD utility, by running command(in 
 project root folder):
 rdmd "-Isrc/" -ofsimple_run_test test/simple_run_test.d

 [...]


I can reproduce this.
The function which fails is:
DirAttrsBlock.this(DirAttrKind attrKind, 
DirValueAttr!true[string] attrs)
Mar 25 2017
parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 17:31:33 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 13:34:43 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 [...]


 If you wish to help. You are welcome) I created branch in my 
 repository `https://github.com/uranuz/ivy` called 
 `v0.1/fail_on_compiler_error`. Seems that error depends on the 
 order of source files that were passed to DMD. The way that 
 error occurs is using RDMD utility, by running command(in 
 project root folder):
 rdmd "-Isrc/" -ofsimple_run_test test/simple_run_test.d

 [...]


 I can reproduce this.
 The function which fails is:
 DirAttrsBlock.this(DirAttrKind attrKind, 
 DirValueAttr!true[string] attrs)


Yes. I fugured out that it fails on this line too. But I don't 
understand why. Compiler fails on assert and doesn't give any 
meaningful info about reasons. In release version of compiler 
it's just segfault. And I don't understand why changing order of 
source files passed to DMD makes it working... I'm not enough 
good in knowlege of compiler internals, so it's surprise for me...
Mar 25 2017
parent reply Stefan Koch <uplink.coder googlemail.com> writes:
On Saturday, 25 March 2017 at 17:41:49 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 17:31:33 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 13:34:43 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 07:57:11 UTC, Stefan Koch wrote:

 [...]


 If you wish to help. You are welcome) I created branch in my 
 repository `https://github.com/uranuz/ivy` called 
 `v0.1/fail_on_compiler_error`. Seems that error depends on 
 the order of source files that were passed to DMD. The way 
 that error occurs is using RDMD utility, by running 
 command(in project root folder):
 rdmd "-Isrc/" -ofsimple_run_test test/simple_run_test.d

 [...]


 I can reproduce this.
 The function which fails is:
 DirAttrsBlock.this(DirAttrKind attrKind, 
 DirValueAttr!true[string] attrs)


 Yes. I fugured out that it fails on this line too. But I don't 
 understand why. Compiler fails on assert and doesn't give any 
 meaningful info about reasons. In release version of compiler 
 it's just segfault. And I don't understand why changing order 
 of source files passed to DMD makes it working... I'm not 
 enough good in knowlege of compiler internals, so it's surprise 
 for me...


It might be that a the particular template does not get 
instantiated if the order is different.
The assert triggers because the constructor is indeed a function 
which returns void.
But at this stage it is not expected to return void.
This is indicative of a compiler bug.
Mar 25 2017
parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 17:47:18 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 17:41:49 UTC, Uranuz wrote:

 On Saturday, 25 March 2017 at 17:31:33 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 13:34:43 UTC, Uranuz wrote:

 [...]


 I can reproduce this.
 The function which fails is:
 DirAttrsBlock.this(DirAttrKind attrKind, 
 DirValueAttr!true[string] attrs)


 Yes. I fugured out that it fails on this line too. But I don't 
 understand why. Compiler fails on assert and doesn't give any 
 meaningful info about reasons. In release version of compiler 
 it's just segfault. And I don't understand why changing order 
 of source files passed to DMD makes it working... I'm not 
 enough good in knowlege of compiler internals, so it's 
 surprise for me...


 It might be that a the particular template does not get 
 instantiated if the order is different.
 The assert triggers because the constructor is indeed a 
 function which returns void.
 But at this stage it is not expected to return void.
 This is indicative of a compiler bug.


What is a good practice to make a bug report if I can't give some 
reduced case when problem occurs? Is it normal to just put link 
to my repo and branch with this problem and put instruction how 
to repeat this bug?
Mar 25 2017
parent reply Stefan Koch <uplink.coder googlemail.com> writes:
On Saturday, 25 March 2017 at 18:10:30 UTC, Uranuz wrote:

 What is a good practice to make a bug report if I can't give 
 some reduced case when problem occurs? Is it normal to just put 
 link to my repo and branch with this problem and put 
 instruction how to repeat this bug?


It is very important to reduce these bugs.
Since we compiler devs like to focus on the problem rather then 
figureing out a certain build process.
I am going to reduce this now.
Mar 25 2017
parent reply Uranuz <neuranuz gmail.com> writes:
On Saturday, 25 March 2017 at 18:19:34 UTC, Stefan Koch wrote:

 On Saturday, 25 March 2017 at 18:10:30 UTC, Uranuz wrote:

 What is a good practice to make a bug report if I can't give 
 some reduced case when problem occurs? Is it normal to just 
 put link to my repo and branch with this problem and put 
 instruction how to repeat this bug?


 It is very important to reduce these bugs.
 Since we compiler devs like to focus on the problem rather then 
 figureing out a certain build process.
 I am going to reduce this now.


Thanks for this) I found more interesting info. In file 
src/ivy/common.d I put local imports:
import ivy.compiler: IDirectiveCompiler;
import ivy.interpreter: INativeDirectiveInterpreter;
...into global scope. And now it says that it conflicts with 
symbols from the place it was imported from. Maybe these 
problemes are connected somehow... or I'm doing something 
completely wrong)

uranuz freelancer:~/sources/ivy$ rdmd "-Isrc/" -ofsimple_run_test 
test/simple_run_test.d
src/ivy/common.d(13): Error: undefined identifier 
'INativeDirectiveInterpreter'
src/ivy/interpreter.d(247): Error: 
ivy.common.INativeDirectiveInterpreter at src/ivy/common.d(13) 
conflicts with ivy.interpreter_data.INativeDirectiveInterpreter 
at src/ivy/interpreter_data.d(336)
src/ivy/interpreter.d(249): Error: function 
ivy.interpreter.RenderDirInterpreter.interpret does not override 
any function
src/ivy/interpreter.d(271): Error: function 
ivy.interpreter.RenderDirInterpreter.attrBlocks does not override 
any function
src/ivy/common.d(13): Error: 
ivy.common.INativeDirectiveInterpreter at src/ivy/common.d(13) 
conflicts with ivy.interpreter_data.INativeDirectiveInterpreter 
at src/ivy/interpreter_data.d(336)
Failed: ["dmd", "-Isrc/", "-v", "-o-", "test/simple_run_test.d", 
"-Itest"]
Mar 25 2017
parent Stefan Koch <uplink.coder googlemail.com> writes:
On Saturday, 25 March 2017 at 18:19:34 UTC, Stefan Koch wrote:

 It is very important to reduce these bugs.
 Since we compiler devs like to focus on the problem rather then 
 figureing out a certain build process.
 I am going to reduce this now.


This seems to be related to an import circle.
a template in interpreter_data.d needs node.d
node.d needs common.d
a struct declaration in common.d needs interpreter_data.d
Mar 25 2017