Merge branch 'rs/mailinfo-qp-decode-fix' into maint
authorJunio C Hamano <gitster@pobox.com>
Wed, 18 Oct 2017 05:19:03 +0000 (14:19 +0900)
committerJunio C Hamano <gitster@pobox.com>
Wed, 18 Oct 2017 05:19:03 +0000 (14:19 +0900)
"git mailinfo" was loose in decoding quoted printable and produced
garbage when the two letters after the equal sign are not
hexadecimal. This has been fixed.

* rs/mailinfo-qp-decode-fix:
mailinfo: don't decode invalid =XY quoted-printable sequences

1  2 
mailinfo.c
diff --combined mailinfo.c
index bd574cb75210334b1a4628f182743b49bd389cd7,5a597ef89ca6f8cfc0affe0ec3659d663f680b4b..70187e3eb30b102c7038a1d1c4e0033d856c99ac
@@@ -1,5 -1,4 +1,5 @@@
  #include "cache.h"
 +#include "config.h"
  #include "utf8.h"
  #include "strbuf.h"
  #include "mailinfo.h"
@@@ -58,17 -57,17 +58,17 @@@ static void parse_bogus_from(struct mai
  static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  {
        int c;
 -      int take_next_litterally = 0;
 +      int take_next_literally = 0;
  
        strbuf_addch(outbuf, '(');
  
        while ((c = *in++) != 0) {
 -              if (take_next_litterally == 1) {
 -                      take_next_litterally = 0;
 +              if (take_next_literally == 1) {
 +                      take_next_literally = 0;
                } else {
                        switch (c) {
                        case '\\':
 -                              take_next_litterally = 1;
 +                              take_next_literally = 1;
                                continue;
                        case '(':
                                in = unquote_comment(outbuf, in);
  static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  {
        int c;
 -      int take_next_litterally = 0;
 +      int take_next_literally = 0;
  
        while ((c = *in++) != 0) {
 -              if (take_next_litterally == 1) {
 -                      take_next_litterally = 0;
 +              if (take_next_literally == 1) {
 +                      take_next_literally = 0;
                } else {
                        switch (c) {
                        case '\\':
 -                              take_next_litterally = 1;
 +                              take_next_literally = 1;
                                continue;
                        case '"':
                                return in;
@@@ -368,11 -367,16 +368,16 @@@ static struct strbuf *decode_q_segment(
  
        while ((c = *in++) != 0) {
                if (c == '=') {
-                       int d = *in++;
+                       int ch, d = *in;
                        if (d == '\n' || !d)
                                break; /* drop trailing newline */
-                       strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
-                       continue;
+                       ch = hex2chr(in);
+                       if (ch >= 0) {
+                               strbuf_addch(out, ch);
+                               in += 2;
+                               continue;
+                       }
+                       /* garbage -- fall through */
                }
                if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
                        c = 0x20;
@@@ -578,26 -582,26 +583,26 @@@ static int check_header(struct mailinf
                goto check_header_out;
        }
  
 -      /* for inbody stuff */
 -      if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
 -              ret = is_format_patch_separator(line->buf + 1, line->len - 1);
 -              goto check_header_out;
 -      }
 -      if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 -              for (i = 0; header[i]; i++) {
 -                      if (!strcmp("Subject", header[i])) {
 -                              handle_header(&hdr_data[i], line);
 -                              ret = 1;
 -                              goto check_header_out;
 -                      }
 -              }
 -      }
 -
  check_header_out:
        strbuf_release(&sb);
        return ret;
  }
  
 +/*
 + * Returns 1 if the given line or any line beginning with the given line is an
 + * in-body header (that is, check_header will succeed when passed
 + * mi->s_hdr_data).
 + */
 +static int is_inbody_header(const struct mailinfo *mi,
 +                          const struct strbuf *line)
 +{
 +      int i;
 +      for (i = 0; header[i]; i++)
 +              if (!mi->s_hdr_data[i] && cmp_header(line, header[i]))
 +                      return 1;
 +      return 0;
 +}
 +
  static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
  {
        struct strbuf *ret;
@@@ -655,35 -659,37 +660,35 @@@ static inline int patchbreak(const stru
        return 0;
  }
  
 -static int is_scissors_line(const struct strbuf *line)
 +static int is_scissors_line(const char *line)
  {
 -      size_t i, len = line->len;
 +      const char *c;
        int scissors = 0, gap = 0;
 -      int first_nonblank = -1;
 -      int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 -      const char *buf = line->buf;
 +      const char *first_nonblank = NULL, *last_nonblank = NULL;
 +      int visible, perforation = 0, in_perforation = 0;
  
 -      for (i = 0; i < len; i++) {
 -              if (isspace(buf[i])) {
 +      for (c = line; *c; c++) {
 +              if (isspace(*c)) {
                        if (in_perforation) {
                                perforation++;
                                gap++;
                        }
                        continue;
                }
 -              last_nonblank = i;
 -              if (first_nonblank < 0)
 -                      first_nonblank = i;
 -              if (buf[i] == '-') {
 +              last_nonblank = c;
 +              if (first_nonblank == NULL)
 +                      first_nonblank = c;
 +              if (*c == '-') {
                        in_perforation = 1;
                        perforation++;
                        continue;
                }
 -              if (i + 1 < len &&
 -                  (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 -                   !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 +              if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 +                   !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
                        in_perforation = 1;
                        perforation += 2;
                        scissors += 2;
 -                      i++;
 +                      c++;
                        continue;
                }
                in_perforation = 0;
         * than half of the perforation.
         */
  
 -      visible = last_nonblank - first_nonblank + 1;
 +      if (first_nonblank && last_nonblank)
 +              visible = last_nonblank - first_nonblank + 1;
 +      else
 +              visible = 0;
        return (scissors && 8 <= visible &&
                visible < perforation * 3 &&
                gap * 2 < perforation);
  }
  
 +static void flush_inbody_header_accum(struct mailinfo *mi)
 +{
 +      if (!mi->inbody_header_accum.len)
 +              return;
 +      if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 +              die("BUG: inbody_header_accum, if not empty, must always contain a valid in-body header");
 +      strbuf_reset(&mi->inbody_header_accum);
 +}
 +
 +static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 +{
 +      if (mi->inbody_header_accum.len &&
 +          (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 +              if (mi->use_scissors && is_scissors_line(line->buf)) {
 +                      /*
 +                       * This is a scissors line; do not consider this line
 +                       * as a header continuation line.
 +                       */
 +                      flush_inbody_header_accum(mi);
 +                      return 0;
 +              }
 +              strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 +              strbuf_addbuf(&mi->inbody_header_accum, line);
 +              return 1;
 +      }
 +
 +      flush_inbody_header_accum(mi);
 +
 +      if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 +              return is_format_patch_separator(line->buf + 1, line->len - 1);
 +      if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 +              int i;
 +              for (i = 0; header[i]; i++)
 +                      if (!strcmp("Subject", header[i])) {
 +                              handle_header(&mi->s_hdr_data[i], line);
 +                              return 1;
 +                      }
 +              return 0;
 +      }
 +      if (is_inbody_header(mi, line)) {
 +              strbuf_addbuf(&mi->inbody_header_accum, line);
 +              return 1;
 +      }
 +      return 0;
 +}
 +
  static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
  {
        assert(!mi->filter_stage);
  
        if (mi->header_stage) {
 -              if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 +              if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 +                      if (mi->inbody_header_accum.len) {
 +                              flush_inbody_header_accum(mi);
 +                              mi->header_stage = 0;
 +                      }
                        return 0;
 +              }
        }
  
        if (mi->use_inbody_headers && mi->header_stage) {
 -              mi->header_stage = check_header(mi, line, mi->s_hdr_data, 0);
 +              mi->header_stage = check_inbody_header(mi, line);
                if (mi->header_stage)
                        return 0;
        } else
        if (convert_to_utf8(mi, line, mi->charset.buf))
                return 0; /* mi->input_error already set */
  
 -      if (mi->use_scissors && is_scissors_line(line)) {
 +      if (mi->use_scissors && is_scissors_line(line->buf)) {
                int i;
  
                strbuf_setlen(&mi->log_message, 0);
@@@ -883,10 -835,7 +888,10 @@@ static int read_one_header_line(struct 
        for (;;) {
                int peek;
  
 -              peek = fgetc(in); ungetc(peek, in);
 +              peek = fgetc(in);
 +              if (peek == EOF)
 +                      break;
 +              ungetc(peek, in);
                if (peek != ' ' && peek != '\t')
                        break;
                if (strbuf_getline_lf(&continuation, in))
@@@ -920,7 -869,8 +925,7 @@@ again
                /* we hit an end boundary */
                /* pop the current boundary off the stack */
                strbuf_release(*(mi->content_top));
 -              free(*(mi->content_top));
 -              *(mi->content_top) = NULL;
 +              FREE_AND_NULL(*(mi->content_top));
  
                /* technically won't happen as is_multipart_boundary()
                   will fail first.  But just in case..
@@@ -1023,8 -973,6 +1028,8 @@@ static void handle_body(struct mailinf
                        break;
        } while (!strbuf_getwholeline(line, mi->input, '\n'));
  
 +      flush_inbody_header_accum(mi);
 +
  handle_body_out:
        strbuf_release(&prev);
  }
@@@ -1102,10 -1050,6 +1107,10 @@@ int mailinfo(struct mailinfo *mi, cons
  
        do {
                peek = fgetc(mi->input);
 +              if (peek == EOF) {
 +                      fclose(cmitmsg);
 +                      return error("empty patch: '%s'", patch);
 +              }
        } while (isspace(peek));
        ungetc(peek, mi->input);
  
@@@ -1144,7 -1088,6 +1149,7 @@@ void setup_mailinfo(struct mailinfo *mi
        strbuf_init(&mi->email, 0);
        strbuf_init(&mi->charset, 0);
        strbuf_init(&mi->log_message, 0);
 +      strbuf_init(&mi->inbody_header_accum, 0);
        mi->header_stage = 1;
        mi->use_inbody_headers = 1;
        mi->content_top = mi->content;
@@@ -1158,7 -1101,6 +1163,7 @@@ void clear_mailinfo(struct mailinfo *mi
        strbuf_release(&mi->name);
        strbuf_release(&mi->email);
        strbuf_release(&mi->charset);
 +      strbuf_release(&mi->inbody_header_accum);
        free(mi->message_id);
  
        for (i = 0; mi->p_hdr_data[i]; i++)