digitalmars.D.learn - Spawning a pty in D
- Colin Grogan (41/41) Oct 17 2013 Im having an issue
- Adam D. Ruppe (20/25) Oct 17 2013 I actually have been writing a terminal emulator for the last few
- Colin Grogan (14/40) Oct 22 2013 Thanks for that Adam, was a great help to me.
- Adam D. Ruppe (9/9) Oct 22 2013 looking at the comments:
Im having an issue I can link to the C header file pty.h (http://man7.org/linux/man-pages/man3/openpty.3.html) and call forkpty like here: http://dpaste.dzfl.pl/c3b07855 You have to compile that by linking with the util library ( add "libs-posix": ["util"] to dubs package.json ) For ease of reference, the main function is: import std.stdio; import pty; void main(){ int master, slave; char[16] name; int pid = forkpty(&master, &name[0], null, null); writefln("PID: %s", pid); writefln("Master : %s", master); writefln("Filename: %s", name); } If I run that I get: ~/Projects/D/dexpect$ ./dexpect PID: 2224 Master : 3 Filename: /dev/pts/6����� (As you can see, I'm attempting to create a D version of the expect library, good way to learn some of the intricacies of D and will be useful I suspect) What I get back is a file name and a File Descriptor to the master side of the pty. How do I then open this file and perform IO on it, preferably in a D way (std.stdio.File hopefully) rather than through C's mess of function calls? If I open /dev/pts/6 with File("/dev/pts/6", "rw") and attempt to write to it, I get a "Bad file descriptor" error. Reading from that file blocks (which I think is correct...) Anyone have any experience with this? As an aside, I'd prefer to do this in a pure D way, and not have to compile against any external C libraries, does anyone know if it is possible to spawn a pty in D without resorting to calling external C libs? Sorry for the long post! Thanks
Oct 17 2013
On Thursday, 17 October 2013 at 13:53:39 UTC, Colin Grogan wrote:Anyone have any experience with this?I actually have been writing a terminal emulator for the last few weeks https://github.com/adamdruppe/terminal-emulator But for reading and writing from the pty, I just used the unix read and write syscalls instead of wrapping it. tbh I don't think there's much point in wrapping it; I think std.stdio.File is hard to use for any byte-at-a-time tasks anyway... If you do want to wrap it... well I think you'd have to modify phobos. The constructor that takes a FILE* is private. (std.stdio is itself just a wrapper around C's <stdio.h>) But I wouldn't even bother, it is easiest to just use "import core.sys.posix.unistd;" and then read()/write() to it.As an aside, I'd prefer to do this in a pure D way, and not have to compile against any external C libraries, does anyone know if it is possible to spawn a pty in D without resorting to calling external C libs?eh you could probably open /dev/ptmx and the other /dev/pts/* to test and reimplement what openpty does yourself, but there really is no pure D way, because it is perfectly normal in D to use C interfaces to talk to the operating system (it IS possible to use D without a C lib, but even druntime assumes it is there). I've never seen a unix install without the terminal util lib, so it is basically part of the OS.
Oct 17 2013
On Thursday, 17 October 2013 at 14:12:36 UTC, Adam D. Ruppe wrote:On Thursday, 17 October 2013 at 13:53:39 UTC, Colin Grogan wrote:Thanks for that Adam, was a great help to me. I studied your code quite a bit (and reused some of it if that's ok?!). I stuck an initial draft of dexpect up on github, located: https://github.com/grogancolin/dexpect if you want to see the fruits of your labor :) Theres still some bugs I need to iron out, but I threw it up there to keep safe nonetheless. I ended up going with what you said and didnt wrap any of the C functions, turns out using them is kind of satisfying and pretty easy anyway. Just need to read up on documentation a bit more is all! Cheers, ColinAnyone have any experience with this?I actually have been writing a terminal emulator for the last few weeks https://github.com/adamdruppe/terminal-emulator But for reading and writing from the pty, I just used the unix read and write syscalls instead of wrapping it. tbh I don't think there's much point in wrapping it; I think std.stdio.File is hard to use for any byte-at-a-time tasks anyway... If you do want to wrap it... well I think you'd have to modify phobos. The constructor that takes a FILE* is private. (std.stdio is itself just a wrapper around C's <stdio.h>) But I wouldn't even bother, it is easiest to just use "import core.sys.posix.unistd;" and then read()/write() to it.As an aside, I'd prefer to do this in a pure D way, and not have to compile against any external C libraries, does anyone know if it is possible to spawn a pty in D without resorting to calling external C libs?eh you could probably open /dev/ptmx and the other /dev/pts/* to test and reimplement what openpty does yourself, but there really is no pure D way, because it is perfectly normal in D to use C interfaces to talk to the operating system (it IS possible to use D without a C lib, but even druntime assumes it is there). I've never seen a unix install without the terminal util lib, so it is basically part of the OS.
Oct 22 2013
looking at the comments: //pragma(lib, "util"); // pragma does not work for me at moment. TODO: FIND OUT WHY! If you are using gdc, and I think ldc too, that pragma is unsupported since it doesn't fit well into the gcc code - there's no easy way to pass linker flags from the D frontend to the gcc backend. So it is a dmd only thing right now. Otherwise, pretty cool. Feel free to use anything of mine you want!
Oct 22 2013