• Alexey Prokhin <alexey.prokhin yandex.ru> Feb 23 2011
• Trass3r <un known.com> Feb 23 2011
• Don <nospam nospam.com> Feb 25 2011
• Jacob Carlborg <doob me.com> Feb 25 2011
• Kagamin <spam here.lot> Feb 25 2011
• dsimcha <dsimcha yahoo.com> Feb 23 2011
• Alexey Prokhin <alexey.prokhin yandex.ru> Feb 23 2011
• Andrew Wiley <debio264 gmail.com> Feb 24 2011
• Michel Fortin <michel.fortin michelf.com> Feb 24 2011

Alexey Prokhin <alexey.prokhin yandex.ru> writes:

 LDC2 Status


I currently work on linux X86_64 port: ldc2 already compiles druntime and
phobos, although there are some serious issues with GC (which make ldc2 quite
useless on 64 bits).

Trass3r Wrote:
 Also I got the feeling that LDC2 has been hastily updated to the newest dmd
frontend without verifying each upgrade step by step and now things are subtly
broken.




I need a hand with writing a good building instruction. If you want to help,
send me a letter, I'll talk you through building process.
By the way, druntime and phobos with ldc2 patches can be found here:
https://github.com/AlexeyProkhin.

Trass3r <un known.com> writes:

 Trass3r Wrote:
 Also I got the feeling that LDC2 has been hastily updated to the newest  
 dmd frontend without verifying each upgrade step by step and now things  
 are subtly broken.


 even compile). So, yes, I decided to update ldc to the latest dmd  
 frontend at the time. But after that every frontend merge was thoroughly  
 tested using phobos unittests and dmd test suite.


Glad to hear that! Didn't mean to talk your efforts down ;)
Hope you can make it usable. Now if only those LLVM suckers implemented  
SEH...


 I need a hand with writing a good building instruction. If you want to  
 help, send me a letter, I'll talk you through building process.
 By the way, druntime and phobos with ldc2 patches can be found here:  
 https://github.com/AlexeyProkhin.


I might try compiling it on Windows again.

Don <nospam nospam.com> writes:

Trass3r wrote:
 Trass3r Wrote:
 Also I got the feeling that LDC2 has been hastily updated to the 
 newest dmd frontend without verifying each upgrade step by step and 
 now things are subtly broken.


 even compile). So, yes, I decided to update ldc to the latest dmd 
 frontend at the time. But after that every frontend merge was 
 thoroughly tested using phobos unittests and dmd test suite.


 Glad to hear that! Didn't mean to talk your efforts down ;)
 Hope you can make it usable. Now if only those LLVM suckers implemented 
 SEH...


Wouldn't it be possible for LDC to use the exception handling 
implementation in druntime? (With some modifications, of course). With 
the last compiler release, there is no longer any dependency on the DMC 
runtime. I put a lot of effort into understanding Windows SEH, and I 
believe it is quite well documented now.


 I need a hand with writing a good building instruction. If you want to 
 help, send me a letter, I'll talk you through building process.
 By the way, druntime and phobos with ldc2 patches can be found here: 
 https://github.com/AlexeyProkhin.


 I might try compiling it on Windows again.

Jacob Carlborg <doob me.com> writes:

On 2011-02-25 10:10, Don wrote:
 Trass3r wrote:
 Trass3r Wrote:
 Also I got the feeling that LDC2 has been hastily updated to the
 newest dmd frontend without verifying each upgrade step by step and
 now things are subtly broken.


 even compile). So, yes, I decided to update ldc to the latest dmd
 frontend at the time. But after that every frontend merge was
 thoroughly tested using phobos unittests and dmd test suite.


 Glad to hear that! Didn't mean to talk your efforts down ;)
 Hope you can make it usable. Now if only those LLVM suckers
 implemented SEH...


 Wouldn't it be possible for LDC to use the exception handling
 implementation in druntime? (With some modifications, of course). With
 the last compiler release, there is no longer any dependency on the DMC
 runtime. I put a lot of effort into understanding Windows SEH, and I
 believe it is quite well documented now.


I think that the compiler outputs exception tables, or something like 
that. Without does tables no exceptions.

 I need a hand with writing a good building instruction. If you want
 to help, send me a letter, I'll talk you through building process.
 By the way, druntime and phobos with ldc2 patches can be found here:
 https://github.com/AlexeyProkhin.


 I might try compiling it on Windows again.





-- 
/Jacob Carlborg

Kagamin <spam here.lot> writes:

Don Wrote:

 Wouldn't it be possible for LDC to use the exception handling 
 implementation in druntime? (With some modifications, of course). With 
 the last compiler release, there is no longer any dependency on the DMC 
 runtime. I put a lot of effort into understanding Windows SEH, and I 
 believe it is quite well documented now.


The runtime is not a problem, llvm lacks compile time support for SEH. They may
try to get sjlj to work.

dsimcha <dsimcha yahoo.com> writes:

On 2/23/2011 3:00 PM, Alexey Prokhin wrote:
 LDC2 Status


 I currently work on linux X86_64 port: ldc2 already compiles druntime and
phobos, although there are some serious issues with GC (which make ldc2 quite
useless on 64 bits).

 Trass3r Wrote:
 Also I got the feeling that LDC2 has been hastily updated to the newest dmd
frontend without verifying each upgrade step by step and now things are subtly
broken.




 I need a hand with writing a good building instruction. If you want to help,
send me a letter, I'll talk you through building process.
 By the way, druntime and phobos with ldc2 patches can be found here:
https://github.com/AlexeyProkhin.


Still can't get druntime to compile.  The readme file in runtime/ is 
horribly outdated.  I cloned your github repo, and tried a few things, 
but I have no clue where I should put it, or how to build it.  It looks 
like the make file is still for DMD.

Alexey Prokhin <alexey.prokhin yandex.ru> writes:

 Still can't get druntime to compile.  The readme file in runtime/ is
 horribly outdated. I cloned your github repo, and tried a few things,
 but I have no clue where I should put it, or how to build it.  It looks
 like the make file is still for DMD.


$ cd ldc2_build_dir
$ cmake -DRUNTIME_DIR=path_to_druntime -DPHOBOS2_DIR=path_to_phobos .
$ make phobos2 (or 'make runtime' if you want to build only druntime)

Andrew Wiley <debio264 gmail.com> writes:

On Thu, Feb 24, 2011 at 12:56 AM, Alexey Prokhin
<alexey.prokhin yandex.ru> wrote:
 I use cmake to compile druntime and phobos:
 $ cd ldc2_build_dir
 $ cmake -DRUNTIME_DIR=path_to_druntime -DPHOBOS2_DIR=path_to_phobos .
 $ make phobos2 (or 'make runtime' if you want to build only druntime)


I tried building on ARM just for the heck of it, and ldc2 compiles
successfully, but when I compile druntime, odd things happen.
Specifically, looking at the version blocks in core/stdc/stdargs.d, I
should be getting a static assert failure because my platform is
unsupported, but instead I'm seeing this:

$ ldc2  core/stdc/stdarg.d
core/stdc/stdio.d(345): Error: undefined identifier va_list
core/stdc/stdio.d(345): Error: va_list is used as a type
core/stdc/stdio.d(346): Error: undefined identifier va_list
core/stdc/stdio.d(346): Error: va_list is used as a type
core/stdc/stdio.d(347): Error: undefined identifier va_list
core/stdc/stdio.d(347): Error: va_list is used as a type
core/stdc/stdio.d(348): Error: undefined identifier va_list
core/stdc/stdio.d(348): Error: va_list is used as a type
core/stdc/stdio.d(349): Error: undefined identifier va_list
core/stdc/stdio.d(349): Error: va_list is used as a type
core/stdc/stdio.d(350): Error: undefined identifier va_list
core/stdc/stdio.d(350): Error: va_list is used as a type
core/stdc/stdio.d(405): Error: undefined identifier va_list
core/stdc/stdio.d(405): Error: va_list is used as a type
core/stdc/wchar_.d(37): Error: undefined identifier va_list
core/stdc/wchar_.d(37): Error: va_list is used as a type
core/stdc/wchar_.d(38): Error: undefined identifier va_list
core/stdc/wchar_.d(38): Error: va_list is used as a type
core/stdc/wchar_.d(39): Error: undefined identifier va_list
core/stdc/wchar_.d(39): Error: va_list is used as a type
core/stdc/wchar_.d(40): Error: undefined identifier va_list
core/stdc/wchar_.d(40): Error: va_list is used as a type
core/stdc/wchar_.d(41): Error: undefined identifier va_list
core/stdc/wchar_.d(41): Error: va_list is used as a type
core/stdc/wchar_.d(42): Error: undefined identifier va_list
core/stdc/wchar_.d(42): Error: va_list is used as a type


This is odd because stdarg contains no import statements. Is the
compiler generating references to the other modules somehow, and if
so, why do I get these errors instead of the unsupported platform
error?

Michel Fortin <michel.fortin michelf.com> writes:

On 2011-02-24 10:49:31 -0500, Andrew Wiley <debio264 gmail.com> said:

 On Thu, Feb 24, 2011 at 12:56 AM, Alexey Prokhin
 <alexey.prokhin yandex.ru> wrote:
 I use cmake to compile druntime and phobos:
 $ cd ldc2_build_dir
 $ cmake -DRUNTIME_DIR=path_to_druntime -DPHOBOS2_DIR=path_to_phobos .
 $ make phobos2 (or 'make runtime' if you want to build only druntime)
 


 I tried building on ARM just for the heck of it, and ldc2 compiles
 successfully, but when I compile druntime, odd things happen.
 Specifically, looking at the version blocks in core/stdc/stdargs.d, I
 should be getting a static assert failure because my platform is
 unsupported, but instead I'm seeing this:
 
 $ ldc2  core/stdc/stdarg.d
 core/stdc/stdio.d(345): Error: undefined identifier va_list
 core/stdc/stdio.d(345): Error: va_list is used as a type
 core/stdc/stdio.d(346): Error: undefined identifier va_list
 core/stdc/stdio.d(346): Error: va_list is used as a type
 core/stdc/stdio.d(347): Error: undefined identifier va_list
 core/stdc/stdio.d(347): Error: va_list is used as a type
 core/stdc/stdio.d(348): Error: undefined identifier va_list
 core/stdc/stdio.d(348): Error: va_list is used as a type
 core/stdc/stdio.d(349): Error: undefined identifier va_list
 core/stdc/stdio.d(349): Error: va_list is used as a type
 core/stdc/stdio.d(350): Error: undefined identifier va_list
 core/stdc/stdio.d(350): Error: va_list is used as a type
 core/stdc/stdio.d(405): Error: undefined identifier va_list
 core/stdc/stdio.d(405): Error: va_list is used as a type
 core/stdc/wchar_.d(37): Error: undefined identifier va_list
 core/stdc/wchar_.d(37): Error: va_list is used as a type
 core/stdc/wchar_.d(38): Error: undefined identifier va_list
 core/stdc/wchar_.d(38): Error: va_list is used as a type
 core/stdc/wchar_.d(39): Error: undefined identifier va_list
 core/stdc/wchar_.d(39): Error: va_list is used as a type
 core/stdc/wchar_.d(40): Error: undefined identifier va_list
 core/stdc/wchar_.d(40): Error: va_list is used as a type
 core/stdc/wchar_.d(41): Error: undefined identifier va_list
 core/stdc/wchar_.d(41): Error: va_list is used as a type
 core/stdc/wchar_.d(42): Error: undefined identifier va_list
 core/stdc/wchar_.d(42): Error: va_list is used as a type
 
 
 This is odd because stdarg contains no import statements. Is the
 compiler generating references to the other modules somehow, and if
 so, why do I get these errors instead of the unsupported platform
 error?


The object module in druntime is always imported implicitly. You could 
try commenting everything in module object and reenabling things as you 
need them.


-- 
Michel Fortin
michel.fortin michelf.com
http://michelf.com/