Re: DMD different compiler behaviour on Linux and Windows
On Friday, 26 April 2019 at 15:48:51 UTC, Ron Tarrant wrote: On Thursday, 25 April 2019 at 20:38:31 UTC, Mike Parker wrote: If you compile with -m32 on Windows the error goes away. Not trying to be a but it also works with -m64 on Windows. Yes, thanks. That's a typo. -m32, where size_t is uint, is the default. In -m64, size_t is ulong.
Re: DMD different compiler behaviour on Linux and Windows
On Thursday, 25 April 2019 at 20:38:31 UTC, Mike Parker wrote: If you compile with -m32 on Windows the error goes away. Not trying to be a but it also works with -m64 on Windows.
Re: DMD different compiler behaviour on Linux and Windows
On Thursday, 25 April 2019 at 20:18:28 UTC, Zans wrote:
import std.stdio;
void main()
{
char[] mychars;
mychars ~= 'a';
long index = 0L;
writeln(mychars[index]);
}
Why would the code above compile perfectly on Linux (Ubuntu
16.04), however it would produce the following error on Windows
10:
source\app.d(8,21): Error: cannot implicitly convert expression
index of type long to uint
On both operating systems DMD version is 2.085.0.
DMD defaults to 64-bit output on 64-Bit Linux but always to
32-bit output on Windows, If you compile with -m32 on Windows the
error goes away. Reasons::
* Array indices are default typed as size_t, which is uint in
32-bit and ulong in 64. long is not implicitly convertible to
uint.
* On Linux, both 64- and 32-bit builds of DMD are available and
the output for each defaults to the same. Only the 32-bit build
is distributed on Windows.
* Historically, compiling 64-bit binaries on Windows required a
separate installation of the Microsoft Build Tools ( or Visual
Studio) and/or the Windows SDK. If 64-bit output were the
default, DMD would not work out of the box. Recently, DMD has
been shipping with the lld linker and some MinGW-based Windows
libraries so that the additional installation is not required and
64-bit compiles can work out of the box, but it’s still
considered experimental.
When out-of-the-box 64-bit compilation is solid and 64-bit builds
are distributed on Windows, the default behavior should be the
same as on Linux.
Re: DMD different compiler behaviour on Linux and Windows
On Thursday, 25 April 2019 at 20:18:28 UTC, Zans wrote:
import std.stdio;
void main()
{
char[] mychars;
mychars ~= 'a';
long index = 0L;
writeln(mychars[index]);
}
Why would the code above compile perfectly on Linux (Ubuntu
16.04), however it would produce the following error on Windows
10:
source\app.d(8,21): Error: cannot implicitly convert expression
index of type long to uint
On both operating systems DMD version is 2.085.0.
The issue here is not Windows vs Linux but 32 bits vs 64 bits.
On 32 bits architectures size_t is defined as uint, long being 64
bits long, conversion from long to uint is a truncating cast
which are not allowed implicitely in D.
It is unfortunate that the D compiler on Windows is still
delivered by default as a 32 bits binary and generating 32 bits
code. I think the next release will start to deliver the compiler
as 64 bits binary and generating 64 bits code.
DMD different compiler behaviour on Linux and Windows
import std.stdio;
void main()
{
char[] mychars;
mychars ~= 'a';
long index = 0L;
writeln(mychars[index]);
}
Why would the code above compile perfectly on Linux (Ubuntu
16.04), however it would produce the following error on Windows
10:
source\app.d(8,21): Error: cannot implicitly convert expression
index of type long to uint
On both operating systems DMD version is 2.085.0.
