# [PATCH 1/2] GIT: Try all addresses for given remote name 5 messages from 2005-07-21 to 2005-07-23. Participants: YOSHIFUJI Hideaki / 吉藤英明, Petr Baudis, Sam Ravnborg. Thread: https://gitlist.dev/t/1285 ## YOSHIFUJI Hideaki / 吉藤英明, 2005-07-21 13:10 Subject: [PATCH 1/2] GIT: Try all addresses for given remote name Message-ID: <20050721.091036.01119516.yoshfuji@linux-ipv6.org> URL: https://gitlist.dev/e/20050721.091036.01119516.yoshfuji%40linux-ipv6.org ``` Hello. Try all addresses for given remote name until it succeeds. Also supports IPv6. Signed-of-by: Hideaki YOSHIFUJI diff --git a/connect.c b/connect.c --- a/connect.c +++ b/connect.c @@ -96,42 +96,57 @@ static enum protocol get_protocol(const die("I don't handle protocol '%s'", name); } -static void lookup_host(const char *host, struct sockaddr *in) -{ - struct addrinfo *res; - int ret; - - ret = getaddrinfo(host, NULL, NULL, &res); - if (ret) - die("Unable to look up %s (%s)", host, gai_strerror(ret)); - *in = *res->ai_addr; - freeaddrinfo(res); -} +#define STR_(s) # s +#define STR(s) STR_(s) static int git_tcp_connect(int fd[2], const char *prog, char *host, char *path) { - struct sockaddr addr; - int port = DEFAULT_GIT_PORT, sockfd; - char *colon; - - colon = strchr(host, ':'); - if (colon) { - char *end; - unsigned long n = strtoul(colon+1, &end, 0); - if (colon[1] && !*end) { - *colon = 0; - port = n; + int sockfd = -1; + char *colon, *end; + char *port = STR(DEFAULT_GIT_PORT); + struct addrinfo hints, *ai0, *ai; + int gai; + + if (host[0] == '[') { + end = strchr(host + 1, ']'); + if (end) { + *end = 0; + end++; + host++; + } else + end = host; + } else + end = host; + colon = strchr(end, ':'); + + if (colon) + port = colon + 1; + + memset(&hints, 0, sizeof(hints)); + hints.ai_socktype = SOCK_STREAM; + hints.ai_protocol = IPPROTO_TCP; + + gai = getaddrinfo(host, port, &hints, &ai); + if (gai) + die("Unable to look up %s (%s)", host, gai_strerror(gai)); + + for (ai0 = ai; ai; ai = ai->ai_next) { + sockfd = socket(ai->ai_family, ai->ai_socktype, ai->ai_protocol); + if (sockfd < 0) + continue; + if (connect(sockfd, ai->ai_addr, ai->ai_addrlen) < 0) { + close(sockfd); + sockfd = -1; + continue; } + break; } - lookup_host(host, &addr); - ((struct sockaddr_in *)&addr)->sin_port = htons(port); + freeaddrinfo(ai0); - sockfd = socket(PF_INET, SOCK_STREAM, IPPROTO_IP); if (sockfd < 0) die("unable to create socket (%s)", strerror(errno)); - if (connect(sockfd, (void *)&addr, sizeof(addr)) < 0) - die("unable to connect (%s)", strerror(errno)); + fd[0] = sockfd; fd[1] = sockfd; packet_write(sockfd, "%s %s\n", prog, path); -- YOSHIFUJI Hideaki @ USAGI Project GPG-FP : 9022 65EB 1ECF 3AD1 0BDF 80D8 4807 F894 E062 0EEA ``` ## Petr Baudis, 2005-07-22 21:09 Subject: Re: [PATCH 1/2] GIT: Try all addresses for given remote name Message-ID: <20050722210913.GH11916@pasky.ji.cz> URL: https://gitlist.dev/e/20050722210913.GH11916%40pasky.ji.cz In-Reply-To: <20050721.091036.01119516.yoshfuji@linux-ipv6.org> ``` Dear diary, on Thu, Jul 21, 2005 at 03:10:36PM CEST, I got a letter where "YOSHIFUJI Hideaki / ?$B5HF#1QL@" told me that... > Hello. Hello, > Try all addresses for given remote name until it succeeds. > Also supports IPv6. > > Signed-of-by: Hideaki YOSHIFUJI > > diff --git a/connect.c b/connect.c > --- a/connect.c > +++ b/connect.c > @@ -96,42 +96,57 @@ static enum protocol get_protocol(const > die("I don't handle protocol '%s'", name); > } > > -static void lookup_host(const char *host, struct sockaddr *in) > -{ > - struct addrinfo *res; > - int ret; > - > - ret = getaddrinfo(host, NULL, NULL, &res); > - if (ret) > - die("Unable to look up %s (%s)", host, gai_strerror(ret)); > - *in = *res->ai_addr; > - freeaddrinfo(res); > -} > +#define STR_(s) # s > +#define STR(s) STR_(s) Uh-huh? Why two macros? Well, why any macros at all? > static int git_tcp_connect(int fd[2], const char *prog, char *host, char *path) > { > - struct sockaddr addr; > - int port = DEFAULT_GIT_PORT, sockfd; > - char *colon; > - > - colon = strchr(host, ':'); > - if (colon) { > - char *end; > - unsigned long n = strtoul(colon+1, &end, 0); > - if (colon[1] && !*end) { > - *colon = 0; > - port = n; > + int sockfd = -1; > + char *colon, *end; > + char *port = STR(DEFAULT_GIT_PORT); > + struct addrinfo hints, *ai0, *ai; > + int gai; > + > + if (host[0] == '[') { > + end = strchr(host + 1, ']'); > + if (end) { > + *end = 0; > + end++; > + host++; > + } else > + end = host; > + } else > + end = host; > + colon = strchr(end, ':'); > + > + if (colon) > + port = colon + 1; > + > + memset(&hints, 0, sizeof(hints)); > + hints.ai_socktype = SOCK_STREAM; > + hints.ai_protocol = IPPROTO_TCP; > + > + gai = getaddrinfo(host, port, &hints, &ai); > + if (gai) > + die("Unable to look up %s (%s)", host, gai_strerror(gai)); > + > + for (ai0 = ai; ai; ai = ai->ai_next) { > + sockfd = socket(ai->ai_family, ai->ai_socktype, ai->ai_protocol); > + if (sockfd < 0) > + continue; > + if (connect(sockfd, ai->ai_addr, ai->ai_addrlen) < 0) { > + close(sockfd); > + sockfd = -1; > + continue; > } > + break; > } > > - lookup_host(host, &addr); > - ((struct sockaddr_in *)&addr)->sin_port = htons(port); > + freeaddrinfo(ai0); > > - sockfd = socket(PF_INET, SOCK_STREAM, IPPROTO_IP); > if (sockfd < 0) > die("unable to create socket (%s)", strerror(errno)); > - if (connect(sockfd, (void *)&addr, sizeof(addr)) < 0) > - die("unable to connect (%s)", strerror(errno)); > + > fd[0] = sockfd; > fd[1] = sockfd; > packet_write(sockfd, "%s %s\n", prog, path); You are saying that you were unable to create socket while you just were unable to connect. Not any biggie, but it saves the user the trouble of one strace after being confused by an error message. :-) -- Petr "Pasky" Baudis Stuff: http://pasky.or.cz/ If you want the holes in your knowledge showing up try teaching someone. -- Alan Cox ``` ## YOSHIFUJI Hideaki / 吉藤英明, 2005-07-22 21:26 Subject: Re: [PATCH 1/2] GIT: Try all addresses for given remote name Message-ID: <20050722.172629.73295196.yoshfuji@linux-ipv6.org> URL: https://gitlist.dev/e/20050722.172629.73295196.yoshfuji%40linux-ipv6.org In-Reply-To: <20050722210913.GH11916@pasky.ji.cz> ``` In article <20050722210913.GH11916@pasky.ji.cz> (at Fri, 22 Jul 2005 23:09:13 +0200), Petr Baudis says: > > -} > > +#define STR_(s) # s > > +#define STR(s) STR_(s) > > Uh-huh? Why two macros? Well, why any macros at all? > : > > + char *colon, *end; > > + char *port = STR(DEFAULT_GIT_PORT); > > + struct addrinfo hints, *ai0, *ai; The macro is used here. This is trick. After preprocess, /* --- cut here --- */ #define TEST 12345 #define STR_(s) # s #define STR(s) STR_(s) test(STR(TEST)); test(STR_(TEST)); test(# TEST); /* --- cut here --- */ becomes test("12345"); test("TEST"); test(# 12345); > > if (sockfd < 0) > > die("unable to create socket (%s)", strerror(errno)); > > - if (connect(sockfd, (void *)&addr, sizeof(addr)) < 0) > > - die("unable to connect (%s)", strerror(errno)); : > You are saying that you were unable to create socket while you just were > unable to connect. Not any biggie, but it saves the user the trouble of > one strace after being confused by an error message. :-) In fact, I don't think it is really worng, because it says that it could not create (connected) socket or endpoint of connection. Anyway, I agree that it would be confusing. Better ideas / wordings? --yoshfuji ``` ## Petr Baudis, 2005-07-22 21:37 Subject: Re: [PATCH 1/2] GIT: Try all addresses for given remote name Message-ID: <20050722213753.GK11916@pasky.ji.cz> URL: https://gitlist.dev/e/20050722213753.GK11916%40pasky.ji.cz In-Reply-To: <20050722.172629.73295196.yoshfuji@linux-ipv6.org> ``` Dear diary, on Fri, Jul 22, 2005 at 11:26:29PM CEST, I got a letter where "YOSHIFUJI Hideaki / ?$B5HF#1QL@" told me that... > In article <20050722210913.GH11916@pasky.ji.cz> (at Fri, 22 Jul 2005 23:09:13 +0200), Petr Baudis says: > > > > -} > > > +#define STR_(s) # s > > > +#define STR(s) STR_(s) > > > > Uh-huh? Why two macros? Well, why any macros at all? > > > : > > > + char *colon, *end; > > > + char *port = STR(DEFAULT_GIT_PORT); > > > + struct addrinfo hints, *ai0, *ai; > > The macro is used here. > This is trick. > > After preprocess, > > /* --- cut here --- */ > #define TEST 12345 > #define STR_(s) # s > #define STR(s) STR_(s) > > test(STR(TEST)); > test(STR_(TEST)); > test(# TEST); > /* --- cut here --- */ > > becomes > > test("12345"); > test("TEST"); > test(# 12345); I see, I didn't know about this. Well, I didn't actually use the # stringificator many times. Perhaps a short comment would be due. > > > if (sockfd < 0) > > > die("unable to create socket (%s)", strerror(errno)); > > > - if (connect(sockfd, (void *)&addr, sizeof(addr)) < 0) > > > - die("unable to connect (%s)", strerror(errno)); > : > > You are saying that you were unable to create socket while you just were > > unable to connect. Not any biggie, but it saves the user the trouble of > > one strace after being confused by an error message. :-) > > In fact, I don't think it is really worng, because it says that > it could not create (connected) socket or endpoint of connection. It uses only two words ("create socket") from that, though. ;-) > Anyway, I agree that it would be confusing. > Better ideas / wordings? if (connect() < 0) sockfd = -2; ... if (sockfd == -1) die("unable to create socket"); else die("unable to connect") or I'd prefer the second error message "unable to connect", which is perhaps more descriptive regarding what effectively happened. BTW, could the close() on the socket ever fail (read as "modify errno") here? -- Petr "Pasky" Baudis Stuff: http://pasky.or.cz/ If you want the holes in your knowledge showing up try teaching someone. -- Alan Cox ``` ## Sam Ravnborg, 2005-07-23 00:15 Subject: Re: [PATCH 1/2] GIT: Try all addresses for given remote name Message-ID: <20050723001550.GB8047@mars.ravnborg.org> URL: https://gitlist.dev/e/20050723001550.GB8047%40mars.ravnborg.org In-Reply-To: <20050722.172629.73295196.yoshfuji@linux-ipv6.org> ``` On Fri, Jul 22, 2005 at 05:26:29PM -0400, YOSHIFUJI Hideaki / ?$B5HF#1QL@ wrote: > In article <20050722210913.GH11916@pasky.ji.cz> (at Fri, 22 Jul 2005 23:09:13 +0200), Petr Baudis says: > > > > -} > > > +#define STR_(s) # s > > > +#define STR(s) STR_(s) > > > > Uh-huh? Why two macros? Well, why any macros at all? > > > : > > > + char *colon, *end; > > > + char *port = STR(DEFAULT_GIT_PORT); > > > + struct addrinfo hints, *ai0, *ai; > > The macro is used here. You could copy the include/linux/stingify.h file from the kernel also. That makes it available for other users also. Sam ```