[Issue 21425] Using va_start twice results in wrong values
https://issues.dlang.org/show_bug.cgi?id=21425 Dlang Bot changed: What|Removed |Added Keywords||pull --- Comment #3 from Dlang Bot --- @WalterBright created dlang/dmd pull request #15267 "fix Issue 21425 - Using va_start twice results in wrong values" fixing this issue: - fix Issue 21425 - Using va_start twice results in wrong values https://github.com/dlang/dmd/pull/15267 --
[Issue 23934] Literal integer multiplication overflow is allowed in CTFE
https://issues.dlang.org/show_bug.cgi?id=23934 Luís Ferreira changed: What|Removed |Added CC||cont...@lsferreira.net Hardware|x86_64 |All OS|Linux |All Severity|enhancement |normal --
[Issue 23934] Literal integer multiplication overflow is allowed in CTFE
https://issues.dlang.org/show_bug.cgi?id=23934 Dlang Bot changed: What|Removed |Added Keywords||pull --- Comment #1 from Dlang Bot --- @ljmf00 created dlang/dmd pull request #15266 "constfold: multiplication should check for overflow" fixing this issue: - constfold: multiplication should check for overflow Fixes issue 23934. Signed-off-by: Luís Ferreira https://github.com/dlang/dmd/pull/15266 --
[Issue 23933] auto return type disables DIP1000 scope check
https://issues.dlang.org/show_bug.cgi?id=23933
Richard Cattermole changed:
What|Removed |Added
Status|NEW |RESOLVED
CC||alphaglosi...@gmail.com
Resolution|--- |INVALID
--- Comment #4 from Richard Cattermole ---
The inference can be seen when you use auto instead of scope inside of
safeUsage when you output the AST.
```d
struct Struct
{
int* p;
@safe void hmm(int i)
{
}
}
```
vs with scope & auto:
```d
struct Struct
{
int* p;
auto pure nothrow @nogc @safe void hmm(int i)
{
}
}
```
I don't see a bug here. This looks to be working correctly. A method that was
not annotated with scope, used with scope, works when inference is turned on.
So I'm closing.
--
[Issue 23935] ImportC for libuv
https://issues.dlang.org/show_bug.cgi?id=23935
--- Comment #4 from a11e99z ---
fixed manually. not checked yet with calling libuv from D, but dmd gives
uv9.obj.
MSVC command:
cl.exe -P -Fiuv.i -DUSING_UV_SHARED -DWIN32_LEAN_AND_MEAN -D__int64="long long"
-D__int32="long" -D__cdecl="" -D__inline="inline" -D__forceinline="inline"
-D__ptr64="" -D__ptr32="" -D__unaligned="" -D_CRT_SECURE_NO_WARNINGS /std:c11
uv.h
and fixing is:
1) changed
typedef struct __declspec(align(16)) __pragma(warning(push))
__pragma(warning(disable:4845)) __declspec(no_init_all) __pragma(warning(pop))
_CONTEXT { ... } CONTEXT, *PCONTEXT;
with issues:
uv9.i(4529): Error: missing tag `identifier` after `struct`
uv9.i(4529): Error: identifier or `(` expected
uv9.i(4598): Error: no type for declarator before `}`
uv9.i(4598): Error: no type-specifier for declarator
to:
typedef struct /*__declspec(align(16)) __pragma(warning(push))
__pragma(warning(disable:4845)) __declspec(no_init_all)
__pragma(warning(pop))*/ _CONTEXT { ... } CONTEXT __attribute__((aligned(
16))), *PCONTEXT;
2) removed '\' (no-line-break here)
with issues:
uv9.i(19384): Error: character '\' is not a valid token
uv9.i(19385): Error: character '\' is not a valid token
(2) should be fixed 100%
(1) hmm, struct attrs should be reordered?
--
[Issue 23935] ImportC for libuv
https://issues.dlang.org/show_bug.cgi?id=23935 --- Comment #3 from a11e99z --- still fixing issues manually hint: '\' means code follows on next line w/o line break todo: ignore it in parser? --
[Issue 23935] ImportC for libuv
https://issues.dlang.org/show_bug.cgi?id=23935 a11e99z changed: What|Removed |Added Attachment #1875|0 |1 is obsolete|| --- Comment #2 from a11e99z --- Created attachment 1876 --> https://issues.dlang.org/attachment.cgi?id=1876&action=edit comressed uv9.i after fixing some issues preprocessed header with command: cl.exe -P -Fiuv.i -DUSING_UV_SHARED -DWIN32_LEAN_AND_MEAN -D__int64="long long" -D__int32="long" -D__cdecl="" -D__inline="inline" -D__forceinline="inline" -D__ptr64="" -D__ptr32="" -D__unaligned="" /std:c11 uv.h --
[Issue 23935] ImportC for libuv
https://issues.dlang.org/show_bug.cgi?id=23935 --- Comment #1 from a11e99z --- so 1st simple step: dmd/importc can not eat unsigned __intXX: typedef unsigned __int64 uintptr_t; gives: Error: missing comma or semicolon after declaration of `__anonymous`, found `uintptr_t` instead --
[Issue 23935] New: ImportC for libuv
https://issues.dlang.org/show_bug.cgi?id=23935 Issue ID: 23935 Summary: ImportC for libuv Product: D Version: D2 Hardware: x86_64 OS: Windows Status: NEW Severity: major Priority: P1 Component: dmd Assignee: nob...@puremagic.com Reporter: blac...@bk.ru Created attachment 1875 --> https://issues.dlang.org/attachment.cgi?id=1875&action=edit proceeded uv9.i 1) I compiled uv.dll with vcpkg 2) I preprocessed "uv.h" (all libuv stuff) with command MSVC: cl.exe -P -Fiuv.i -DUSING_UV_SHARED -DWIN32_LEAN_AND_MEAN uv.h and got uv.i 3) I removed empty/#pragma/#line lines from the last one with commands: @findstr /v /r /c:"^[ \x09]*$" /c:"^$" uv.i >uv2.i @findstr /v /r /c:"^.*#line" uv2.i >uv3.i @findstr /v /r /c:"^.*#pragma warning" /c:"^.*#pragma once" /c:"^.*#pragma region" /c:"^.*#pragma endregion" uv3.i >uv9.i and got uv9.i ~1MB (in attachments) 4) dmd/ldc2 -c uv9.i gives to me many issues: uv9.i(2): Error: missing comma or semicolon after declaration of `__anonymous`, found `uintptr_t` instead ══uv9.i(4): Error: identifier or `(` expected Euv9.i(7): Error: missing comma or semicolon after declaration of `__anonymous`, found `size_t` instead rruv9.i(12): Error: identifier or `(` expected first 12LOC of uv9.i: #pragma pack(push, 8) typedef unsigned __int64 uintptr_t; typedef char* va_list; void __cdecl __va_start(va_list* , ...); #pragma pack(pop) __pragma(pack(push, 8)) typedef unsigned __int64 size_t; typedef __int64 ptrdiff_t; typedef __int64 intptr_t; typedef _Bool __vcrt_bool; typedef unsigned short wchar_t; --
[Issue 23933] auto return type disables DIP1000 scope check
https://issues.dlang.org/show_bug.cgi?id=23933 --- Comment #3 from RazvanN --- So, if we enable attribute inference for all functions that have a body, this will be fixed as a consequence. Or, the workaround would be to manually annotate `hmm` as scope. --
[Issue 23933] auto return type disables DIP1000 scope check
https://issues.dlang.org/show_bug.cgi?id=23933 --- Comment #2 from RazvanN --- [...] for auto functions because a function body needs to be present* --
[Issue 23933] auto return type disables DIP1000 scope check
https://issues.dlang.org/show_bug.cgi?id=23933 RazvanN changed: What|Removed |Added CC||razvan.nitu1...@gmail.com --- Comment #1 from RazvanN --- Under the current specification this is not a bug, even though the behavior is surprising. Attributes are inferred for auto functions because you a function body needs to be present. Attributes are not inferred for non-auto functions, even if the body is present. --
[Issue 23934] New: Literal integer multiplication overflow is allowed in CTFE
https://issues.dlang.org/show_bug.cgi?id=23934 Issue ID: 23934 Summary: Literal integer multiplication overflow is allowed in CTFE Product: D Version: D2 Hardware: x86_64 OS: Linux Status: NEW Severity: enhancement Priority: P1 Component: dmd Assignee: nob...@puremagic.com Reporter: e...@weka.io This example program generates the following assertion failure: static assert(1000*1000*1000*1000 > 0); // Error: static assert: `-727379968 > 0` is false It would detect bugs earlier if constant-folding/CTFE execution of multiplication disallowed overflow. --
[Issue 23933] auto return type disables DIP1000 scope check
https://issues.dlang.org/show_bug.cgi?id=23933 Atila Neves changed: What|Removed |Added Keywords||accepts-invalid, safe --
[Issue 23933] New: auto return type disables DIP1000 scope check
https://issues.dlang.org/show_bug.cgi?id=23933
Issue ID: 23933
Summary: auto return type disables DIP1000 scope check
Product: D
Version: D2
Hardware: x86_64
OS: Linux
Status: NEW
Severity: normal
Priority: P1
Component: dmd
Assignee: nob...@puremagic.com
Reporter: atila.ne...@gmail.com
This fails to compile with -preview=dip1000:
---
struct Struct {
int* p;
void hmm(int i) @safe { }
}
void safeUsage() @safe {
int i;
scope s = Struct();
s.hmm(42);
}
---
bug.d(9): Error: scope variable `s` assigned to non-scope parameter `this`
calling `hmm`
But changing `hmm`'s return type to `auto` makes the error go away.
--
