Reference git-commit-tree for env vars.
[gitweb.git] / mailinfo.c
index 6c4d5898238704954a0b67b99079193f588795b6..3c56f8c10801bdc93eaaac6319e477f81c096717 100644 (file)
@@ -2,19 +2,36 @@
  * Another stupid program, this one parsing the headers of an
  * email to figure out authorship and subject
  */
+#define _GNU_SOURCE
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 #include <ctype.h>
+#ifndef NO_ICONV
+#include <iconv.h>
+#endif
+#include "git-compat-util.h"
+#include "cache.h"
 
 static FILE *cmitmsg, *patchfile;
 
+static int keep_subject = 0;
+static char *metainfo_charset = NULL;
 static char line[1000];
 static char date[1000];
 static char name[1000];
 static char email[1000];
 static char subject[1000];
 
+static enum  {
+       TE_DONTCARE, TE_QP, TE_BASE64,
+} transfer_encoding;
+static char charset[256];
+
+static char multipart_boundary[1000];
+static int multipart_boundary_len;
+static int patch_lines = 0;
+
 static char *sanity_check(char *name, char *email)
 {
        int len = strlen(name);
@@ -25,13 +42,43 @@ static char *sanity_check(char *name, char *email)
        return name;
 }
 
+static int bogus_from(char *line)
+{
+       /* John Doe <johndoe> */
+       char *bra, *ket, *dst, *cp;
+
+       /* This is fallback, so do not bother if we already have an
+        * e-mail address.
+        */ 
+       if (*email)
+               return 0;
+
+       bra = strchr(line, '<');
+       if (!bra)
+               return 0;
+       ket = strchr(bra, '>');
+       if (!ket)
+               return 0;
+
+       for (dst = email, cp = bra+1; cp < ket; )
+               *dst++ = *cp++;
+       *dst = 0;
+       for (cp = line; isspace(*cp); cp++)
+               ;
+       for (bra--; isspace(*bra); bra--)
+               *bra = 0;
+       cp = sanity_check(cp, email);
+       strcpy(name, cp);
+       return 1;
+}
+
 static int handle_from(char *line)
 {
        char *at = strchr(line, '@');
        char *dst;
 
        if (!at)
-               return 0;
+               return bogus_from(line);
 
        /*
         * If we already have one email, don't take any confusing lines
@@ -39,99 +86,195 @@ static int handle_from(char *line)
        if (*email && strchr(at+1, '@'))
                return 0;
 
+       /* Pick up the string around '@', possibly delimited with <>
+        * pair; that is the email part.  White them out while copying.
+        */
        while (at > line) {
                char c = at[-1];
-               if (isspace(c) || c == '<')
+               if (isspace(c))
                        break;
+               if (c == '<') {
+                       at[-1] = ' ';
+                       break;
+               }
                at--;
        }
        dst = email;
        for (;;) {
                unsigned char c = *at;
-               if (!c || c == '>' || isspace(c))
+               if (!c || c == '>' || isspace(c)) {
+                       if (c == '>')
+                               *at = ' ';
                        break;
+               }
                *at++ = ' ';
                *dst++ = c;
        }
        *dst++ = 0;
 
+       /* The remainder is name.  It could be "John Doe <john.doe@xz>"
+        * or "john.doe@xz (John Doe)", but we have whited out the
+        * email part, so trim from both ends, possibly removing
+        * the () pair at the end.
+        */
        at = line + strlen(line);
        while (at > line) {
                unsigned char c = *--at;
-               if (isalnum(c))
+               if (!isspace(c)) {
+                       at[(c == ')') ? 0 : 1] = 0;
                        break;
-               *at = 0;
+               }
        }
 
        at = line;
        for (;;) {
                unsigned char c = *at;
-               if (!c)
-                       break;
-               if (isalnum(c))
+               if (!c || !isspace(c)) {
+                       if (c == '(')
+                               at++;
                        break;
+               }
                at++;
        }
-
        at = sanity_check(at, email);
-       
        strcpy(name, at);
        return 1;
 }
 
-static void handle_date(char *line)
+static int handle_date(char *line)
 {
        strcpy(date, line);
+       return 0;
 }
 
-static void handle_subject(char *line)
+static int handle_subject(char *line)
 {
        strcpy(subject, line);
+       return 0;
 }
 
-static void add_subject_line(char *line)
+/* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
+ * to have enough heuristics to grok MIME encoded patches often found
+ * on our mailing lists.  For example, we do not even treat header lines
+ * case insensitively.
+ */
+
+static int slurp_attr(const char *line, const char *name, char *attr)
 {
-       while (isspace(*line))
-               line++;
-       *--line = ' ';
-       strcat(subject, line);
+       char *ends, *ap = strcasestr(line, name);
+       size_t sz;
+
+       if (!ap) {
+               *attr = 0;
+               return 0;
+       }
+       ap += strlen(name);
+       if (*ap == '"') {
+               ap++;
+               ends = "\"";
+       }
+       else
+               ends = "; \t";
+       sz = strcspn(ap, ends);
+       memcpy(attr, ap, sz);
+       attr[sz] = 0;
+       return 1;
 }
 
-static void check_line(char *line, int len)
+static int handle_subcontent_type(char *line)
 {
-       static int cont = -1;
-       if (!memcmp(line, "From:", 5) && isspace(line[5])) {
-               handle_from(line+6);
-               cont = 0;
-               return;
+       /* We do not want to mess with boundary.  Note that we do not
+        * handle nested multipart.
+        */
+       if (strcasestr(line, "boundary=")) {
+               fprintf(stderr, "Not handling nested multipart message.\n");
+               exit(1);
        }
-       if (!memcmp(line, "Date:", 5) && isspace(line[5])) {
-               handle_date(line+6);
-               cont = 0;
-               return;
+       slurp_attr(line, "charset=", charset);
+       if (*charset) {
+               int i, c;
+               for (i = 0; (c = charset[i]) != 0; i++)
+                       charset[i] = tolower(c);
        }
-       if (!memcmp(line, "Subject:", 8) && isspace(line[8])) {
-               handle_subject(line+9);
-               cont = 1;
-               return;
+       return 0;
+}
+
+static int handle_content_type(char *line)
+{
+       *multipart_boundary = 0;
+       if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
+               memcpy(multipart_boundary, "--", 2);
+               multipart_boundary_len = strlen(multipart_boundary);
        }
-       if (isspace(*line)) {
-               switch (cont) {
-               case 0:
-                       fprintf(stderr, "I don't do 'Date:' or 'From:' line continuations\n");
-                       break;
-               case 1:
-                       add_subject_line(line);
-                       return;
-               default:
-                       break;
+       slurp_attr(line, "charset=", charset);
+       return 0;
+}
+
+static int handle_content_transfer_encoding(char *line)
+{
+       if (strcasestr(line, "base64"))
+               transfer_encoding = TE_BASE64;
+       else if (strcasestr(line, "quoted-printable"))
+               transfer_encoding = TE_QP;
+       else
+               transfer_encoding = TE_DONTCARE;
+       return 0;
+}
+
+static int is_multipart_boundary(const char *line)
+{
+       return (!memcmp(line, multipart_boundary, multipart_boundary_len));
+}
+
+static int eatspace(char *line)
+{
+       int len = strlen(line);
+       while (len > 0 && isspace(line[len-1]))
+               line[--len] = 0;
+       return len;
+}
+
+#define SEEN_FROM 01
+#define SEEN_DATE 02
+#define SEEN_SUBJECT 04
+
+/* First lines of body can have From:, Date:, and Subject: */
+static int handle_inbody_header(int *seen, char *line)
+{
+       if (!memcmp("From:", line, 5) && isspace(line[5])) {
+               if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
+                       *seen |= SEEN_FROM;
+                       return 1;
+               }
+       }
+       if (!memcmp("Date:", line, 5) && isspace(line[5])) {
+               if (!(*seen & SEEN_DATE)) {
+                       handle_date(line+6);
+                       *seen |= SEEN_DATE;
+                       return 1;
                }
        }
-       cont = -1;
+       if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
+               if (!(*seen & SEEN_SUBJECT)) {
+                       handle_subject(line+9);
+                       *seen |= SEEN_SUBJECT;
+                       return 1;
+               }
+       }
+       if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
+               if (!(*seen & SEEN_SUBJECT)) {
+                       handle_subject(line);
+                       *seen |= SEEN_SUBJECT;
+                       return 1;
+               }
+       }
+       return 0;
 }
 
-static char * cleanup_subject(char *subject)
+static char *cleanup_subject(char *subject)
 {
+       if (keep_subject)
+               return subject;
        for (;;) {
                char *p;
                int len, remove;
@@ -181,72 +324,456 @@ static void cleanup_space(char *buf)
        }
 }
 
-static void handle_rest(void)
+typedef int (*header_fn_t)(char *);
+struct header_def {
+       const char *name;
+       header_fn_t func;
+       int namelen;
+};
+
+static void check_header(char *line, int len, struct header_def *header)
+{
+       int i;
+
+       if (header[0].namelen <= 0) {
+               for (i = 0; header[i].name; i++)
+                       header[i].namelen = strlen(header[i].name);
+       }
+       for (i = 0; header[i].name; i++) {
+               int len = header[i].namelen;
+               if (!strncasecmp(line, header[i].name, len) &&
+                   line[len] == ':' && isspace(line[len + 1])) {
+                       header[i].func(line + len + 2);
+                       break;
+               }
+       }
+}
+
+static void check_subheader_line(char *line, int len)
+{
+       static struct header_def header[] = {
+               { "Content-Type", handle_subcontent_type },
+               { "Content-Transfer-Encoding",
+                 handle_content_transfer_encoding },
+               { NULL },
+       };
+       check_header(line, len, header);
+}
+static void check_header_line(char *line, int len)
+{
+       static struct header_def header[] = {
+               { "From", handle_from },
+               { "Date", handle_date },
+               { "Subject", handle_subject },
+               { "Content-Type", handle_content_type },
+               { "Content-Transfer-Encoding",
+                 handle_content_transfer_encoding },
+               { NULL },
+       };
+       check_header(line, len, header);
+}
+
+static int read_one_header_line(char *line, int sz, FILE *in)
+{
+       int ofs = 0;
+       while (ofs < sz) {
+               int peek, len;
+               if (fgets(line + ofs, sz - ofs, in) == NULL)
+                       return ofs;
+               len = eatspace(line + ofs);
+               if (len == 0)
+                       return ofs;
+               peek = fgetc(in); ungetc(peek, in);
+               if (peek == ' ' || peek == '\t') {
+                       /* Yuck, 2822 header "folding" */
+                       ofs += len;
+                       continue;
+               }
+               return ofs + len;
+       }
+       return ofs;
+}
+
+static unsigned hexval(int c)
+{
+       if (c >= '0' && c <= '9')
+               return c - '0';
+       if (c >= 'a' && c <= 'f')
+               return c - 'a' + 10;
+       if (c >= 'A' && c <= 'F')
+               return c - 'A' + 10;
+       return ~0;
+}
+
+static int decode_q_segment(char *in, char *ot, char *ep)
+{
+       int c;
+       while ((c = *in++) != 0 && (in <= ep)) {
+               if (c == '=') {
+                       int d = *in++;
+                       if (d == '\n' || !d)
+                               break; /* drop trailing newline */
+                       *ot++ = ((hexval(d) << 4) | hexval(*in++));
+               }
+               else
+                       *ot++ = c;
+       }
+       *ot = 0;
+       return 0;
+}
+
+static int decode_b_segment(char *in, char *ot, char *ep)
+{
+       /* Decode in..ep, possibly in-place to ot */
+       int c, pos = 0, acc = 0;
+
+       while ((c = *in++) != 0 && (in <= ep)) {
+               if (c == '+')
+                       c = 62;
+               else if (c == '/')
+                       c = 63;
+               else if ('A' <= c && c <= 'Z')
+                       c -= 'A';
+               else if ('a' <= c && c <= 'z')
+                       c -= 'a' - 26;
+               else if ('0' <= c && c <= '9')
+                       c -= '0' - 52;
+               else if (c == '=') {
+                       /* padding is almost like (c == 0), except we do
+                        * not output NUL resulting only from it;
+                        * for now we just trust the data.
+                        */
+                       c = 0;
+               }
+               else
+                       continue; /* garbage */
+               switch (pos++) {
+               case 0:
+                       acc = (c << 2);
+                       break;
+               case 1:
+                       *ot++ = (acc | (c >> 4));
+                       acc = (c & 15) << 4;
+                       break;
+               case 2:
+                       *ot++ = (acc | (c >> 2));
+                       acc = (c & 3) << 6;
+                       break;
+               case 3:
+                       *ot++ = (acc | c);
+                       acc = pos = 0;
+                       break;
+               }
+       }
+       *ot = 0;
+       return 0;
+}
+
+static void convert_to_utf8(char *line, char *charset)
+{
+#ifndef NO_ICONV
+       char *in, *out;
+       size_t insize, outsize, nrc;
+       char outbuf[4096]; /* cheat */
+       static char latin_one[] = "latin1";
+       char *input_charset = *charset ? charset : latin_one;
+       iconv_t conv = iconv_open(metainfo_charset, input_charset);
+
+       if (conv == (iconv_t) -1) {
+               static int warned_latin1_once = 0;
+               if (input_charset != latin_one) {
+                       fprintf(stderr, "cannot convert from %s to %s\n",
+                               input_charset, metainfo_charset);
+                       *charset = 0;
+               }
+               else if (!warned_latin1_once) {
+                       warned_latin1_once = 1;
+                       fprintf(stderr, "tried to convert from %s to %s, "
+                               "but your iconv does not work with it.\n",
+                               input_charset, metainfo_charset);
+               }
+               return;
+       }
+       in = line;
+       insize = strlen(in);
+       out = outbuf;
+       outsize = sizeof(outbuf);
+       nrc = iconv(conv, &in, &insize, &out, &outsize);
+       iconv_close(conv);
+       if (nrc == (size_t) -1)
+               return;
+       *out = 0;
+       strcpy(line, outbuf);
+#endif
+}
+
+static void decode_header_bq(char *it)
+{
+       char *in, *out, *ep, *cp, *sp;
+       char outbuf[1000];
+
+       in = it;
+       out = outbuf;
+       while ((ep = strstr(in, "=?")) != NULL) {
+               int sz, encoding;
+               char charset_q[256], piecebuf[256];
+               if (in != ep) {
+                       sz = ep - in;
+                       memcpy(out, in, sz);
+                       out += sz;
+                       in += sz;
+               }
+               /* E.g.
+                * ep : "=?iso-2022-jp?B?GyR...?= foo"
+                * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
+                */
+               ep += 2;
+               cp = strchr(ep, '?');
+               if (!cp)
+                       return; /* no munging */
+               for (sp = ep; sp < cp; sp++)
+                       charset_q[sp - ep] = tolower(*sp);
+               charset_q[cp - ep] = 0;
+               encoding = cp[1];
+               if (!encoding || cp[2] != '?')
+                       return; /* no munging */
+               ep = strstr(cp + 3, "?=");
+               if (!ep)
+                       return; /* no munging */
+               switch (tolower(encoding)) {
+               default:
+                       return; /* no munging */
+               case 'b':
+                       sz = decode_b_segment(cp + 3, piecebuf, ep);
+                       break;
+               case 'q':
+                       sz = decode_q_segment(cp + 3, piecebuf, ep);
+                       break;
+               }
+               if (sz < 0)
+                       return;
+               if (metainfo_charset)
+                       convert_to_utf8(piecebuf, charset_q);
+               strcpy(out, piecebuf);
+               out += strlen(out);
+               in = ep + 2;
+       }
+       strcpy(out, in);
+       strcpy(it, outbuf);
+}
+
+static void decode_transfer_encoding(char *line)
+{
+       char *ep;
+
+       switch (transfer_encoding) {
+       case TE_QP:
+               ep = line + strlen(line);
+               decode_q_segment(line, line, ep);
+               break;
+       case TE_BASE64:
+               ep = line + strlen(line);
+               decode_b_segment(line, line, ep);
+               break;
+       case TE_DONTCARE:
+               break;
+       }
+}
+
+static void handle_info(void)
 {
-       char *sub = cleanup_subject(subject);
+       char *sub;
+       static int done_info = 0;
+
+       if (done_info)
+               return;
+
+       done_info = 1;
+       sub = cleanup_subject(subject);
        cleanup_space(name);
        cleanup_space(date);
        cleanup_space(email);
        cleanup_space(sub);
-       printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n", name, email, sub, date);
-       FILE *out = cmitmsg;
 
+       /* Unwrap inline B and Q encoding, and optionally
+        * normalize the meta information to utf8.
+        */
+       decode_header_bq(name);
+       decode_header_bq(date);
+       decode_header_bq(email);
+       decode_header_bq(sub);
+       printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
+              name, email, sub, date);
+}
+
+/* We are inside message body and have read line[] already.
+ * Spit out the commit log.
+ */
+static int handle_commit_msg(void)
+{
+       if (!cmitmsg)
+               return 0;
        do {
                if (!memcmp("diff -", line, 6) ||
-                   !memcmp("---", line, 3))
-                       out = patchfile;
+                   !memcmp("---", line, 3) ||
+                   !memcmp("Index: ", line, 7))
+                       break;
+               if ((multipart_boundary[0] && is_multipart_boundary(line))) {
+                       /* We come here when the first part had only
+                        * the commit message without any patch.  We
+                        * pretend we have not seen this line yet, and
+                        * go back to the loop.
+                        */
+                       return 1;
+               }
 
-               fputs(line, out);
+               /* Unwrap transfer encoding and optionally
+                * normalize the log message to UTF-8.
+                */
+               decode_transfer_encoding(line);
+               if (metainfo_charset)
+                       convert_to_utf8(line, charset);
+               fputs(line, cmitmsg);
        } while (fgets(line, sizeof(line), stdin) != NULL);
+       fclose(cmitmsg);
+       cmitmsg = NULL;
+       return 0;
+}
 
-       if (out == cmitmsg) {
-               fprintf(stderr, "No patch found\n");
-               exit(1);
-       }
+/* We have done the commit message and have the first
+ * line of the patch in line[].
+ */
+static void handle_patch(void)
+{
+       do {
+               if (multipart_boundary[0] && is_multipart_boundary(line))
+                       break;
+               /* Only unwrap transfer encoding but otherwise do not
+                * do anything.  We do *NOT* want UTF-8 conversion
+                * here; we are dealing with the user payload.
+                */
+               decode_transfer_encoding(line);
+               fputs(line, patchfile);
+               patch_lines++;
+       } while (fgets(line, sizeof(line), stdin) != NULL);
+}
 
-       fclose(cmitmsg);
-       fclose(patchfile);
+/* multipart boundary and transfer encoding are set up for us, and we
+ * are at the end of the sub header.  do equivalent of handle_body up
+ * to the next boundary without closing patchfile --- we will expect
+ * that the first part to contain commit message and a patch, and
+ * handle other parts as pure patches.
+ */
+static int handle_multipart_one_part(void)
+{
+       int seen = 0;
+       int n = 0;
+       int len;
+
+       while (fgets(line, sizeof(line), stdin) != NULL) {
+       again:
+               len = eatspace(line);
+               n++;
+               if (!len)
+                       continue;
+               if (is_multipart_boundary(line))
+                       break;
+               if (0 <= seen && handle_inbody_header(&seen, line))
+                       continue;
+               seen = -1; /* no more inbody headers */
+               line[len] = '\n';
+               handle_info();
+               if (handle_commit_msg())
+                       goto again;
+               handle_patch();
+               break;
+       }
+       if (n == 0)
+               return -1;
+       return 0;
 }
 
-static int eatspace(char *line)
+static void handle_multipart_body(void)
 {
-       int len = strlen(line);
-       while (len > 0 && isspace(line[len-1]))
-               line[--len] = 0;
-       return len;
+       int part_num = 0;
+
+       /* Skip up to the first boundary */
+       while (fgets(line, sizeof(line), stdin) != NULL)
+               if (is_multipart_boundary(line)) {
+                       part_num = 1;
+                       break;
+               }
+       if (!part_num)
+               return;
+       /* We are on boundary line.  Start slurping the subhead. */
+       while (1) {
+               int len = read_one_header_line(line, sizeof(line), stdin);
+               if (!len) {
+                       if (handle_multipart_one_part() < 0)
+                               return;
+                       /* Reset per part headers */
+                       transfer_encoding = TE_DONTCARE;
+                       charset[0] = 0;
+               }
+               else
+                       check_subheader_line(line, len);
+       }
+       fclose(patchfile);
+       if (!patch_lines) {
+               fprintf(stderr, "No patch found\n");
+               exit(1);
+       }
 }
 
+/* Non multipart message */
 static void handle_body(void)
 {
-       int has_from = 0;
+       int seen = 0;
 
-       /* First line of body can be a From: */
        while (fgets(line, sizeof(line), stdin) != NULL) {
                int len = eatspace(line);
                if (!len)
                        continue;
-               if (!memcmp("From:", line, 5) && isspace(line[5])) {
-                       if (!has_from && handle_from(line+6)) {
-                               has_from = 1;
-                               continue;
-                       }
-               }
+               if (0 <= seen && handle_inbody_header(&seen, line))
+                       continue;
+               seen = -1; /* no more inbody headers */
                line[len] = '\n';
-               handle_rest();
+               handle_info();
+               handle_commit_msg();
+               handle_patch();
                break;
        }
+       fclose(patchfile);
+       if (!patch_lines) {
+               fprintf(stderr, "No patch found\n");
+               exit(1);
+       }
 }
 
-static void usage(void)
-{
-       fprintf(stderr, "mailinfo msg-file path-file < email\n");
-       exit(1);
-}
+static const char mailinfo_usage[] =
+       "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
 
-int main(int argc, char ** argv)
+int main(int argc, char **argv)
 {
+       /* NEEDSWORK: might want to do the optional .git/ directory
+        * discovery
+        */
+       git_config(git_default_config);
+
+       while (1 < argc && argv[1][0] == '-') {
+               if (!strcmp(argv[1], "-k"))
+                       keep_subject = 1;
+               else if (!strcmp(argv[1], "-u"))
+                       metainfo_charset = git_commit_encoding;
+               else if (!strncmp(argv[1], "--encoding=", 11))
+                       metainfo_charset = argv[1] + 11;
+               else
+                       usage(mailinfo_usage);
+               argc--; argv++;
+       }
+
        if (argc != 3)
-               usage();
+               usage(mailinfo_usage);
        cmitmsg = fopen(argv[1], "w");
        if (!cmitmsg) {
                perror(argv[1]);
@@ -257,13 +784,16 @@ int main(int argc, char ** argv)
                perror(argv[2]);
                exit(1);
        }
-       while (fgets(line, sizeof(line), stdin) != NULL) {
-               int len = eatspace(line);
+       while (1) {
+               int len = read_one_header_line(line, sizeof(line), stdin);
                if (!len) {
-                       handle_body();
+                       if (multipart_boundary[0])
+                               handle_multipart_body();
+                       else
+                               handle_body();
                        break;
                }
-               check_line(line, len);
+               check_header_line(line, len);
        }
        return 0;
 }