X-Git-Url: https://tinc-vpn.org/git/browse?a=blobdiff_plain;f=src%2Ftincctl.c;h=799da0aa192105cee261392d58c53bf502f8df96;hb=afb175873e6aa10d2d4dca3572edf054968c538d;hp=fc21d42a586a4205d741416504ac40233e5e4202;hpb=aec82bb1c94af6d3142cdef0c51f42f38e9be3e0;p=tinc diff --git a/src/tincctl.c b/src/tincctl.c index fc21d42a..799da0aa 100644 --- a/src/tincctl.c +++ b/src/tincctl.c @@ -810,16 +810,31 @@ static int cmd_start(int argc, char *argv[]) { int nargc = 0; char **nargv = xzalloc((optind + argc) * sizeof *nargv); - nargv[nargc++] = c; + char *arg0 = c; +#ifdef HAVE_MINGW + /* + Windows has no real concept of an "argv array". A command line is just one string. + The CRT of the new process will decode the command line string to generate argv before calling main(), and (by convention) + it uses quotes to handle spaces in arguments. + Therefore we need to quote all arguments that might contain spaces. No, execvp() won't do that for us (see MSDN). + If we don't do that, then execvp() will run fine but any spaces in the filename contained in arg0 will bleed + into the next arguments when the spawned process' CRT parses its command line, resulting in chaos. + */ + xasprintf(&arg0, "\"%s\"", arg0); +#endif + nargv[nargc++] = arg0; for(int i = 1; i < optind; i++) nargv[nargc++] = orig_argv[i]; for(int i = 1; i < argc; i++) nargv[nargc++] = argv[i]; #ifdef HAVE_MINGW - execvp(c, nargv); - fprintf(stderr, "Error starting %s: %s\n", c, strerror(errno)); - return 1; + int status = spawnvp(_P_WAIT, c, nargv); + if (status == -1) { + fprintf(stderr, "Error starting %s: %s\n", c, strerror(errno)); + return 1; + } + return status; #else pid_t pid = fork(); if(pid == -1) {