Remove unnecessary debugging from builtin-fetch
[gitweb.git] / convert.c
index ad106ef35fb0729709b243d6292980f15cd1b0a5..d77c8eb8b2802d675b320ddeb063c1cf70cc57d8 100644 (file)
--- a/convert.c
+++ b/convert.c
@@ -1,5 +1,6 @@
 #include "cache.h"
 #include "attr.h"
+#include "run-command.h"
 
 /*
  * convert.c - convert a file when checking it out and checking it in.
@@ -85,7 +86,7 @@ static char *crlf_to_git(const char *path, const char *src, unsigned long *sizep
        unsigned long size, nsize;
        struct text_stat stats;
 
-       if ((action == CRLF_BINARY) || (action == CRLF_GUESS && !auto_crlf))
+       if ((action == CRLF_BINARY) || !auto_crlf)
                return NULL;
 
        size = *sizep;
@@ -153,7 +154,7 @@ static char *crlf_to_worktree(const char *path, const char *src, unsigned long *
        unsigned char last;
 
        if ((action == CRLF_BINARY) || (action == CRLF_INPUT) ||
-           (action == CRLF_GUESS && auto_crlf <= 0))
+           auto_crlf <= 0)
                return NULL;
 
        size = *sizep;
@@ -200,13 +201,385 @@ static char *crlf_to_worktree(const char *path, const char *src, unsigned long *
        return buffer;
 }
 
+static int filter_buffer(const char *path, const char *src,
+                        unsigned long size, const char *cmd)
+{
+       /*
+        * Spawn cmd and feed the buffer contents through its stdin.
+        */
+       struct child_process child_process;
+       int pipe_feed[2];
+       int write_err, status;
+
+       memset(&child_process, 0, sizeof(child_process));
+
+       if (pipe(pipe_feed) < 0) {
+               error("cannot create pipe to run external filter %s", cmd);
+               return 1;
+       }
+
+       child_process.pid = fork();
+       if (child_process.pid < 0) {
+               error("cannot fork to run external filter %s", cmd);
+               close(pipe_feed[0]);
+               close(pipe_feed[1]);
+               return 1;
+       }
+       if (!child_process.pid) {
+               dup2(pipe_feed[0], 0);
+               close(pipe_feed[0]);
+               close(pipe_feed[1]);
+               execlp("sh", "sh", "-c", cmd, NULL);
+               return 1;
+       }
+       close(pipe_feed[0]);
+
+       write_err = (write_in_full(pipe_feed[1], src, size) < 0);
+       if (close(pipe_feed[1]))
+               write_err = 1;
+       if (write_err)
+               error("cannot feed the input to external filter %s", cmd);
+
+       status = finish_command(&child_process);
+       if (status)
+               error("external filter %s failed %d", cmd, -status);
+       return (write_err || status);
+}
+
+static char *apply_filter(const char *path, const char *src,
+                         unsigned long *sizep, const char *cmd)
+{
+       /*
+        * Create a pipeline to have the command filter the buffer's
+        * contents.
+        *
+        * (child --> cmd) --> us
+        */
+       const int SLOP = 4096;
+       int pipe_feed[2];
+       int status;
+       char *dst;
+       unsigned long dstsize, dstalloc;
+       struct child_process child_process;
+
+       if (!cmd)
+               return NULL;
+
+       memset(&child_process, 0, sizeof(child_process));
+
+       if (pipe(pipe_feed) < 0) {
+               error("cannot create pipe to run external filter %s", cmd);
+               return NULL;
+       }
+
+       fflush(NULL);
+       child_process.pid = fork();
+       if (child_process.pid < 0) {
+               error("cannot fork to run external filter %s", cmd);
+               close(pipe_feed[0]);
+               close(pipe_feed[1]);
+               return NULL;
+       }
+       if (!child_process.pid) {
+               dup2(pipe_feed[1], 1);
+               close(pipe_feed[0]);
+               close(pipe_feed[1]);
+               exit(filter_buffer(path, src, *sizep, cmd));
+       }
+       close(pipe_feed[1]);
+
+       dstalloc = *sizep;
+       dst = xmalloc(dstalloc);
+       dstsize = 0;
+
+       while (1) {
+               ssize_t numread = xread(pipe_feed[0], dst + dstsize,
+                                       dstalloc - dstsize);
+
+               if (numread <= 0) {
+                       if (!numread)
+                               break;
+                       error("read from external filter %s failed", cmd);
+                       free(dst);
+                       dst = NULL;
+                       break;
+               }
+               dstsize += numread;
+               if (dstalloc <= dstsize + SLOP) {
+                       dstalloc = dstsize + SLOP;
+                       dst = xrealloc(dst, dstalloc);
+               }
+       }
+       if (close(pipe_feed[0])) {
+               error("read from external filter %s failed", cmd);
+               free(dst);
+               dst = NULL;
+       }
+
+       status = finish_command(&child_process);
+       if (status) {
+               error("external filter %s failed %d", cmd, -status);
+               free(dst);
+               dst = NULL;
+       }
+
+       if (dst)
+               *sizep = dstsize;
+       return dst;
+}
+
+static struct convert_driver {
+       const char *name;
+       struct convert_driver *next;
+       char *smudge;
+       char *clean;
+} *user_convert, **user_convert_tail;
+
+static int read_convert_config(const char *var, const char *value)
+{
+       const char *ep, *name;
+       int namelen;
+       struct convert_driver *drv;
+
+       /*
+        * External conversion drivers are configured using
+        * "filter.<name>.variable".
+        */
+       if (prefixcmp(var, "filter.") || (ep = strrchr(var, '.')) == var + 6)
+               return 0;
+       name = var + 7;
+       namelen = ep - name;
+       for (drv = user_convert; drv; drv = drv->next)
+               if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
+                       break;
+       if (!drv) {
+               char *namebuf;
+               drv = xcalloc(1, sizeof(struct convert_driver));
+               namebuf = xmalloc(namelen + 1);
+               memcpy(namebuf, name, namelen);
+               namebuf[namelen] = 0;
+               drv->name = namebuf;
+               drv->next = NULL;
+               *user_convert_tail = drv;
+               user_convert_tail = &(drv->next);
+       }
+
+       ep++;
+
+       /*
+        * filter.<name>.smudge and filter.<name>.clean specifies
+        * the command line:
+        *
+        *      command-line
+        *
+        * The command-line will not be interpolated in any way.
+        */
+
+       if (!strcmp("smudge", ep)) {
+               if (!value)
+                       return error("%s: lacks value", var);
+               drv->smudge = strdup(value);
+               return 0;
+       }
+
+       if (!strcmp("clean", ep)) {
+               if (!value)
+                       return error("%s: lacks value", var);
+               drv->clean = strdup(value);
+               return 0;
+       }
+       return 0;
+}
+
 static void setup_convert_check(struct git_attr_check *check)
 {
        static struct git_attr *attr_crlf;
+       static struct git_attr *attr_ident;
+       static struct git_attr *attr_filter;
 
-       if (!attr_crlf)
+       if (!attr_crlf) {
                attr_crlf = git_attr("crlf", 4);
-       check->attr = attr_crlf;
+               attr_ident = git_attr("ident", 5);
+               attr_filter = git_attr("filter", 6);
+               user_convert_tail = &user_convert;
+               git_config(read_convert_config);
+       }
+       check[0].attr = attr_crlf;
+       check[1].attr = attr_ident;
+       check[2].attr = attr_filter;
+}
+
+static int count_ident(const char *cp, unsigned long size)
+{
+       /*
+        * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
+        */
+       int cnt = 0;
+       char ch;
+
+       while (size) {
+               ch = *cp++;
+               size--;
+               if (ch != '$')
+                       continue;
+               if (size < 3)
+                       break;
+               if (memcmp("Id", cp, 2))
+                       continue;
+               ch = cp[2];
+               cp += 3;
+               size -= 3;
+               if (ch == '$')
+                       cnt++; /* $Id$ */
+               if (ch != ':')
+                       continue;
+
+               /*
+                * "$Id: ... "; scan up to the closing dollar sign and discard.
+                */
+               while (size) {
+                       ch = *cp++;
+                       size--;
+                       if (ch == '$') {
+                               cnt++;
+                               break;
+                       }
+               }
+       }
+       return cnt;
+}
+
+static char *ident_to_git(const char *path, const char *src, unsigned long *sizep, int ident)
+{
+       int cnt;
+       unsigned long size;
+       char *dst, *buf;
+
+       if (!ident)
+               return NULL;
+       size = *sizep;
+       cnt = count_ident(src, size);
+       if (!cnt)
+               return NULL;
+       buf = xmalloc(size);
+
+       for (dst = buf; size; size--) {
+               char ch = *src++;
+               *dst++ = ch;
+               if ((ch == '$') && (3 <= size) &&
+                   !memcmp("Id:", src, 3)) {
+                       unsigned long rem = size - 3;
+                       const char *cp = src + 3;
+                       do {
+                               ch = *cp++;
+                               if (ch == '$')
+                                       break;
+                               rem--;
+                       } while (rem);
+                       if (!rem)
+                               continue;
+                       memcpy(dst, "Id$", 3);
+                       dst += 3;
+                       size -= (cp - src);
+                       src = cp;
+               }
+       }
+
+       *sizep = dst - buf;
+       return buf;
+}
+
+static char *ident_to_worktree(const char *path, const char *src, unsigned long *sizep, int ident)
+{
+       int cnt;
+       unsigned long size;
+       char *dst, *buf;
+       unsigned char sha1[20];
+
+       if (!ident)
+               return NULL;
+
+       size = *sizep;
+       cnt = count_ident(src, size);
+       if (!cnt)
+               return NULL;
+
+       hash_sha1_file(src, size, "blob", sha1);
+       buf = xmalloc(size + cnt * 43);
+
+       for (dst = buf; size; size--) {
+               const char *cp;
+               /* Fetch next source character, move the pointer on */
+               char ch = *src++;
+               /* Copy the current character to the destination */
+               *dst++ = ch;
+               /* If the current character is "$" or there are less than three
+                * remaining bytes or the two bytes following this one are not
+                * "Id", then simply read the next character */
+               if ((ch != '$') || (size < 3) || memcmp("Id", src, 2))
+                       continue;
+               /*
+                * Here when
+                *  - There are more than 2 bytes remaining
+                *  - The current three bytes are "$Id"
+                * with
+                *  - ch == "$"
+                *  - src[0] == "I"
+                */
+
+               /*
+                * It's possible that an expanded Id has crept its way into the
+                * repository, we cope with that by stripping the expansion out
+                */
+               if (src[2] == ':') {
+                       /* Expanded keywords have "$Id:" at the front */
+
+                       /* discard up to but not including the closing $ */
+                       unsigned long rem = size - 3;
+                       /* Point at first byte after the ":" */
+                       cp = src + 3;
+                       /*
+                        * Throw away characters until either
+                        *  - we reach a "$"
+                        *  - we run out of bytes (rem == 0)
+                        */
+                       do {
+                               ch = *cp;
+                               if (ch == '$')
+                                       break;
+                               cp++;
+                               rem--;
+                       } while (rem);
+                       /* If the above finished because it ran out of characters, then
+                        * this is an incomplete keyword, so don't run the expansion */
+                       if (!rem)
+                               continue;
+               } else if (src[2] == '$')
+                       cp = src + 2;
+               else
+                       /* Anything other than "$Id:XXX$" or $Id$ and we skip the
+                        * expansion */
+                       continue;
+
+               /* cp is now pointing at the last $ of the keyword */
+
+               memcpy(dst, "Id: ", 4);
+               dst += 4;
+               memcpy(dst, sha1_to_hex(sha1), 40);
+               dst += 40;
+               *dst++ = ' ';
+
+               /* Adjust for the characters we've discarded */
+               size -= (cp - src);
+               src = cp;
+
+               /* Copy the final "$" */
+               *dst++ = *src++;
+               size--;
+       }
+
+       *sizep = dst - buf;
+       return buf;
 }
 
 static int git_path_check_crlf(const char *path, struct git_attr_check *check)
@@ -224,26 +597,93 @@ static int git_path_check_crlf(const char *path, struct git_attr_check *check)
        return CRLF_GUESS;
 }
 
+static struct convert_driver *git_path_check_convert(const char *path,
+                                            struct git_attr_check *check)
+{
+       const char *value = check->value;
+       struct convert_driver *drv;
+
+       if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
+               return NULL;
+       for (drv = user_convert; drv; drv = drv->next)
+               if (!strcmp(value, drv->name))
+                       return drv;
+       return NULL;
+}
+
+static int git_path_check_ident(const char *path, struct git_attr_check *check)
+{
+       const char *value = check->value;
+
+       return !!ATTR_TRUE(value);
+}
+
 char *convert_to_git(const char *path, const char *src, unsigned long *sizep)
 {
-       struct git_attr_check check[1];
+       struct git_attr_check check[3];
        int crlf = CRLF_GUESS;
+       int ident = 0;
+       char *filter = NULL;
+       char *buf, *buf2;
 
        setup_convert_check(check);
-       if (!git_checkattr(path, 1, check)) {
-               crlf = git_path_check_crlf(path, check);
+       if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
+               struct convert_driver *drv;
+               crlf = git_path_check_crlf(path, check + 0);
+               ident = git_path_check_ident(path, check + 1);
+               drv = git_path_check_convert(path, check + 2);
+               if (drv && drv->clean)
+                       filter = drv->clean;
+       }
+
+       buf = apply_filter(path, src, sizep, filter);
+
+       buf2 = crlf_to_git(path, buf ? buf : src, sizep, crlf);
+       if (buf2) {
+               free(buf);
+               buf = buf2;
+       }
+
+       buf2 = ident_to_git(path, buf ? buf : src, sizep, ident);
+       if (buf2) {
+               free(buf);
+               buf = buf2;
        }
-       return crlf_to_git(path, src, sizep, crlf);
+
+       return buf;
 }
 
 char *convert_to_working_tree(const char *path, const char *src, unsigned long *sizep)
 {
-       struct git_attr_check check[1];
+       struct git_attr_check check[3];
        int crlf = CRLF_GUESS;
+       int ident = 0;
+       char *filter = NULL;
+       char *buf, *buf2;
 
        setup_convert_check(check);
-       if (!git_checkattr(path, 1, check)) {
-               crlf = git_path_check_crlf(path, check);
+       if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
+               struct convert_driver *drv;
+               crlf = git_path_check_crlf(path, check + 0);
+               ident = git_path_check_ident(path, check + 1);
+               drv = git_path_check_convert(path, check + 2);
+               if (drv && drv->smudge)
+                       filter = drv->smudge;
+       }
+
+       buf = ident_to_worktree(path, src, sizep, ident);
+
+       buf2 = crlf_to_worktree(path, buf ? buf : src, sizep, crlf);
+       if (buf2) {
+               free(buf);
+               buf = buf2;
+       }
+
+       buf2 = apply_filter(path, buf ? buf : src, sizep, filter);
+       if (buf2) {
+               free(buf);
+               buf = buf2;
        }
-       return crlf_to_worktree(path, src, sizep, crlf);
+
+       return buf;
 }