implement setting of parallel option through server command

src/driver/prove_client.ml

@@ -2,9 +2,6 @@ let standalone : bool ref = ref true
 let socket : Unix.file_descr option ref = ref None
 let max_running_provers : int ref = ref 1
 
-let set_max_running_provers x =
-  max_running_provers := x
-
 let client_connect socket_name =
   if Sys.os_type = "Win32" then begin
     let name = "\\\\.\\pipe\\" ^ socket_name in
@@ -90,9 +87,15 @@ let force_connect () =
       connect()
   | _ -> ()
 
+let set_max_running_provers x =
+  max_running_provers := x;
+  if !socket <> None then
+    send_request_string ("parallel;" ^ string_of_int x)
+
 let send_request ~id ~timelimit ~memlimit ~cmd =
   force_connect ();
   let buf = Buffer.create 128 in
+  Buffer.add_string buf "run;";
   Buffer.add_string buf (string_of_int id);
   Buffer.add_char buf ';';
   Buffer.add_string buf (string_of_int timelimit);

src/server/README.server

@@ -17,10 +17,17 @@ Command line options
 
 A client request is a single line which looks like this:
 
-  id;timeout;memlimit;cmd;arg1;arg2;...;argn
+  commandkind;payload
 
-All items are separated by semicolons, and must not contain semicolons
-themselves (but may contain spaces). Their meaning is the following:
+Where commandkind is a simple string and payload is a semicolon-separated
+list. There are currently only two possible commands. The first command looks
+like this:
+
+  run;id;timeout;memlimit;cmd;arg1;arg2;...;argn
+
+So the commandstring is "run". All items are separated by semicolons, and must
+not contain semicolons themselves (but may contain spaces). Their meaning is
+the following:
   id       - a (ideally) unique identifier which identifies the request
   timeout  - the allowed CPU time in seconds for this command;
              this must be number, 0 for unlimited
@@ -29,7 +36,10 @@ themselves (but may contain spaces). Their meaning is the following:
   cmd      - the name of the executable to run
   arg(i)   - the commandline arguments of the command to run
 
-A server answer looks like this:
+The server does not acknowledge the receipt of this message. However, it will
+run the executable with its arguments and time/memory limit. When the
+executable has terminated, the server sends a message like this to the client
+who sent the 'run' request:
 
   id;exitcode;time;timeout;file
 
@@ -42,6 +52,15 @@ Their meaning is the following:
   file     - the path to a file which contains the stdout and stderr of the
              executed program
 
+The second commmand is like this:
+
+  parallel;num
+
+So the commandstring is "parallel". 'num' is a number greater or equal to 1.
+When this command is received, the server will allow to run up to 'num'
+processes in parallel from now on. Later 'parallel' commands can increase or
+decrease this number. There is no server answer to this command.
+
 There are two separate implementations on linux and windows, but both are
 very similar in structure and share some code (but should share much more).
 They are both essentially a single-threaded event loop, where the possible

src/server/request.c

@@ -2,6 +2,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include "request.h"
+#include "options.h"
 
 //count the semicolons in <buf>, up to <len>
 int count_semicolons(char* buf, int len);
@@ -43,16 +44,37 @@ int copy_up_to_semicolon(char* buf, int begin, int len, char** result) {
 }
 
 prequest parse_request(char* str_req, int len, int key) {
-  int numargs;
+  int numargs, semic, parallel_arg;
   int i = 0;
   int pos = 0;
   prequest req;
   char* tmp;
 
-  numargs = count_semicolons(str_req, len) - 3;
+  semic = count_semicolons(str_req, len);
+  if (semic == 0) {
+    return NULL;
+  }
+  //  might be a 'parallel' command
+  if (semic == 1) {
+    pos = copy_up_to_semicolon (str_req, pos, len, &tmp);
+    if (strncmp(tmp, "parallel", pos) == 0) {
+        pos = copy_up_to_semicolon (str_req, pos, len, &tmp);
+        parallel_arg = atoi(tmp);
+        if (parallel_arg >= 1) {
+          parallel = parallel_arg;
+        }
+    }
+    return NULL;
+  }
+
+  numargs = semic - 4;
   if (numargs < 0) {
     return NULL;
   }
+  pos = copy_up_to_semicolon(str_req, pos, len, &tmp);
+  if (strncmp(tmp, "run", pos) != 0) {
+    return NULL;
+  }
   req = (prequest) malloc(sizeof(request));
   req->key = key;
   req->numargs = numargs;

src/server/server-win.c

@@ -443,7 +443,7 @@ void accept_client(int key) {
    pclient client = (pclient) malloc(sizeof(t_client));
    client->handle = server_socket->handle;
    client->readbuf = init_readbuf(BUFSIZE);
-   client->writebuf = init_writebuf(parallel);
+   client->writebuf = init_writebuf(16);
    init_connect_data(&(client->read), READOP);
    init_connect_data(&(client->write), WRITEOP);
    free(server_socket);
@@ -564,8 +564,8 @@ void init() {
    strcat(socket_name, basename);
 
    queue = init_queue(100);
-   clients = init_list(parallel);
-   processes = init_list(parallel);
+   clients = init_list(16);
+   processes = init_list(16);
 
    init_logging();
    create_server_socket();