Installation on (64 bit) Linux
I managed to get a working system-wide dmd installation. It would have been easier if the Ubuntu packages or linux installation instructions actually set things up so the compiler could find the interface definitions, and if the required dependencies on a 64-bit (Ubuntu) system were documented outside the forums (maybe even automatically in the debs). It would also help if installation instructions were linked more prominently, like from the download page, and high on the side-bar. In firefox, at 1280x1024, on a fresh Ubuntu 9.04 installation the Tools:DMD D Compiler link that ultimately leads to installation instructions is off the bottom of the screen on. http://www.digitalmars.com/d/ I'd suggest a nice Getting D link up in the first panel with Overview and the rest. Brief details: dmd.conf, in all distributions, has include paths like -...@p%/../../src/phobos This will not work if dmd.conf is under $(prefix)/bin and the interfaces under $(prefix)/src. The 2.028 deb omits druntime entirely! The linux install instructions do not mention that you should be able to just run the binaries out of an unpacked zip. (presumably the ../../src that causes so much pain everywhere else is meant to get from dmd/linux/bin to dmd/src). On the other hand, the linux install instructions don't mention that you should also copy /dmd/src into system directories as well if you are going for a system-wide install, and that dmd.conf needs editing if you copy it to /etc (stating /etc/../../src/phobos/object.d is fun). The gcc-multilibs stuff you need to compile on a 64 bit system is not documented anywhere (in particular, not included as a dependency of 64-bit debs). In full detail: The ubuntu packages are both broken. The 1.048 package has a bogus dmd.conf that sends the compiler looking for object.d at /usr/src/phobos/ object.d when it was installed to /usr/local/src/phobos/phobos/object.d. Fixing dmd.conf seems to be enough to get it working, after fixing the 64- bit stuff described later. The 2.028 package is completely broken, completely omitting the druntime directory and consequently object.d, as well as having a similarly bogus dnd.conf. The online linux installation instructions at http://www.digitalmars.com/d/2.0/dmd-linux.html#installation neglect to mention the compiler will work if run in-place from an unpacked zip. The closest they come is if you happen to extend your path to run them out of the unpacked zip, in which case all the cruft the other steps direct you to copy into your system directories will be ignored. Assuming you want a system-wide installation, the instructions say nothing about copying the interfaces from dmd/src (and the readmes (as well as the directory name) suggest it just contains source, not critically important interface files the compiler needs to be able to find at runtime), or the need to edit dmd.conf. Once the above is sorted out those of us on 64-bit platforms still need to get the right 32 bit libraries. A fairly mysterious linker error about libpthread eventually suggests a 32-bit glibc is required, which I had. A bit of searching turned up Walter's instructions: http://www.digitalmars.com/pnews/read.php? server=news.digitalmars.comgroup=digitalmars.D.announceartnum=15219 To compile 32 bit programs under ub64: sudo apt-get install gcc-multilib libc6-i386 lib6-dev-i386 To run 32 bit programs: sudo apt-get install gcc-multilib sudo apt-get install g++-multilib
Re: pyd still usable?
On Tue, Apr 28, 2009 at 4:05 AM, Simen Haugen si...@norstat.no wrote: It uses templates for bindings, and if I remember correctly, dmd hits the limit for max variable name length when using templates. I hit the limit somewhere around 8 functions I think. Actually the name limit hasn't been a problem for several months now. DMD Win now MD5-hashes long symbol names so it's now impossible (?) to crash OPTLINK by having long symbol names. However, this has simply fixed one problem to reveal another: OPTLINK will die if there are too many fixups in a single object. Just a hard crash. This happens a lot in Pyd-like code with several classes and methods wrapped. Might work on linux, but I haven't tried. It does. The only weak link is OPTLINK, which is only used by DMD Win.
Re: Get the name of a function and the parameters?
Jarrett Billingsley wrote:
On Mon, Apr 27, 2009 at 6:56 AM, Jacob Carlborg d...@me.com wrote:
I found that this:
void foo (int x, int y)
{
}
void main ()
{
pragma(msg, typeof(foo).stringof);
}
gave this result:
void function(int x, int y)
So now I just have to get the names out of there.
Interesting! I wonder if that changed recently.
I don't think so, I'm using gdc.
Re: D-styled data file
I changed the prototype to:
void get(in char[] varName, ...)
I'm not totally happy with it because I only need one variadic argument and
:
How do I get the .stringof of an variadic type?
_arguments[0].stringof doesn't work :)
How do I mutate the original argument? (ref)
Please tell me if I'm on the totally wrong track here.
---
module ddata.main;
import std.file;
import std.stdio;
import std.string;
void main()
{
char[] filename = `data.dat`;
char[][] file;
try{
file = splitlines( cast(char[])read(filename) );
}
catch{
throw new Exception(Couldn't load : ~ filename);
}
DData file_dd = new DData(file);
int i;
file_dd.get(`i`, i);
}
class DData
{
private char[][] _file;
this(char[][] file)
{
_file = file.dup;
}
void get(in char[] varName, ...)
{
char[] type = _arguments[0].stringof;
writefln (type);
foreach(int i, char[] line; _file)
{
if(line[0..type.length] == type)
{
writefln (i);
break;
}
writefln (`failed`);
}
}
}
---
Re: Get the name of a function and the parameters?
On Tue, Apr 28, 2009 at 3:07 PM, grauzone n...@example.net wrote: I'd like to pass several functions at once. Is there a way to make this variadic? The obvious approach (writing NameOfFunc(alias f...)) fails with a syntax error. Sure, you'd just make it NameOfFunc(f...) and then recursively instantiate, converting one item at a time to its name until f.length == 0. Alternatively, you can write a compile-time map and use NameOfFunc as the mapping predicate.
Re: Get the name of a function and the parameters?
Jarrett Billingsley wrote: On Tue, Apr 28, 2009 at 3:07 PM, grauzone n...@example.net wrote: I'd like to pass several functions at once. Is there a way to make this variadic? The obvious approach (writing NameOfFunc(alias f...)) fails with a syntax error. Sure, you'd just make it NameOfFunc(f...) and then recursively instantiate, converting one item at a time to its name until f.length == 0. Alternatively, you can write a compile-time map and use NameOfFunc as the mapping predicate. That requires f to be a type, which loses you the actual names. And you cannot (last time I checked) have aliases in a tuple. -- Daniel
