files-backend: remove the use of git_path()
[gitweb.git] / sha1_name.c
index bd10700b3018b61da79553839e8b58a738377580..cda9e49b12e7128dd22a8ccbaff437cb7db34d92 100644 (file)
@@ -7,6 +7,7 @@
 #include "refs.h"
 #include "remote.h"
 #include "dir.h"
+#include "sha1-array.h"
 
 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
 
@@ -14,7 +15,7 @@ typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
 
 struct disambiguate_state {
        int len; /* length of prefix in hex chars */
-       char hex_pfx[GIT_SHA1_HEXSZ];
+       char hex_pfx[GIT_SHA1_HEXSZ + 1];
        unsigned char bin_pfx[GIT_SHA1_RAWSZ];
 
        disambiguate_hint_fn fn;
@@ -90,25 +91,18 @@ static void find_short_object_filename(struct disambiguate_state *ds)
                 * alt->name/alt->base while iterating over the
                 * object databases including our own.
                 */
-               const char *objdir = get_object_directory();
-               size_t objdir_len = strlen(objdir);
-               fakeent = xmalloc(st_add3(sizeof(*fakeent), objdir_len, 43));
-               memcpy(fakeent->base, objdir, objdir_len);
-               fakeent->name = fakeent->base + objdir_len + 1;
-               fakeent->name[-1] = '/';
+               fakeent = alloc_alt_odb(get_object_directory());
        }
        fakeent->next = alt_odb_list;
 
        xsnprintf(hex, sizeof(hex), "%.2s", ds->hex_pfx);
        for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
+               struct strbuf *buf = alt_scratch_buf(alt);
                struct dirent *de;
                DIR *dir;
-               /*
-                * every alt_odb struct has 42 extra bytes after the base
-                * for exactly this purpose
-                */
-               xsnprintf(alt->name, 42, "%.2s/", ds->hex_pfx);
-               dir = opendir(alt->base);
+
+               strbuf_addf(buf, "%.2s/", ds->hex_pfx);
+               dir = opendir(buf->buf);
                if (!dir)
                        continue;
 
@@ -282,6 +276,36 @@ static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unuse
        return kind == OBJ_BLOB;
 }
 
+static disambiguate_hint_fn default_disambiguate_hint;
+
+int set_disambiguate_hint_config(const char *var, const char *value)
+{
+       static const struct {
+               const char *name;
+               disambiguate_hint_fn fn;
+       } hints[] = {
+               { "none", NULL },
+               { "commit", disambiguate_commit_only },
+               { "committish", disambiguate_committish_only },
+               { "tree", disambiguate_tree_only },
+               { "treeish", disambiguate_treeish_only },
+               { "blob", disambiguate_blob_only }
+       };
+       int i;
+
+       if (!value)
+               return config_error_nonbool(var);
+
+       for (i = 0; i < ARRAY_SIZE(hints); i++) {
+               if (!strcasecmp(value, hints[i].name)) {
+                       default_disambiguate_hint = hints[i].fn;
+                       return 0;
+               }
+       }
+
+       return error("unknown hint type for '%s': %s", var, value);
+}
+
 static int init_object_disambiguation(const char *name, int len,
                                      struct disambiguate_state *ds)
 {
@@ -291,7 +315,6 @@ static int init_object_disambiguation(const char *name, int len,
                return -1;
 
        memset(ds, 0, sizeof(*ds));
-       memset(ds->hex_pfx, 'x', GIT_SHA1_HEXSZ);
 
        for (i = 0; i < len ;i++) {
                unsigned char c = name[i];
@@ -313,10 +336,43 @@ static int init_object_disambiguation(const char *name, int len,
        }
 
        ds->len = len;
+       ds->hex_pfx[len] = '\0';
        prepare_alt_odb();
        return 0;
 }
 
+static int show_ambiguous_object(const unsigned char *sha1, void *data)
+{
+       const struct disambiguate_state *ds = data;
+       struct strbuf desc = STRBUF_INIT;
+       int type;
+
+       if (ds->fn && !ds->fn(sha1, ds->cb_data))
+               return 0;
+
+       type = sha1_object_info(sha1, NULL);
+       if (type == OBJ_COMMIT) {
+               struct commit *commit = lookup_commit(sha1);
+               if (commit) {
+                       struct pretty_print_context pp = {0};
+                       pp.date_mode.type = DATE_SHORT;
+                       format_commit_message(commit, " %ad - %s", &desc, &pp);
+               }
+       } else if (type == OBJ_TAG) {
+               struct tag *tag = lookup_tag(sha1);
+               if (!parse_tag(tag) && tag->tag)
+                       strbuf_addf(&desc, " %s", tag->tag);
+       }
+
+       advise("  %s %s%s",
+              find_unique_abbrev(sha1, DEFAULT_ABBREV),
+              typename(type) ? typename(type) : "unknown type",
+              desc.buf);
+
+       strbuf_release(&desc);
+       return 0;
+}
+
 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
                          unsigned flags)
 {
@@ -340,36 +396,98 @@ static int get_short_sha1(const char *name, int len, unsigned char *sha1,
                ds.fn = disambiguate_treeish_only;
        else if (flags & GET_SHA1_BLOB)
                ds.fn = disambiguate_blob_only;
+       else
+               ds.fn = default_disambiguate_hint;
 
        find_short_object_filename(&ds);
        find_short_packed_object(&ds);
        status = finish_object_disambiguation(&ds, sha1);
 
-       if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
-               return error("short SHA1 %.*s is ambiguous.", ds.len, ds.hex_pfx);
+       if (!quietly && (status == SHORT_NAME_AMBIGUOUS)) {
+               error(_("short SHA1 %s is ambiguous"), ds.hex_pfx);
+
+               /*
+                * We may still have ambiguity if we simply saw a series of
+                * candidates that did not satisfy our hint function. In
+                * that case, we still want to show them, so disable the hint
+                * function entirely.
+                */
+               if (!ds.ambiguous)
+                       ds.fn = NULL;
+
+               advise(_("The candidates are:"));
+               for_each_abbrev(ds.hex_pfx, show_ambiguous_object, &ds);
+       }
+
        return status;
 }
 
+static int collect_ambiguous(const unsigned char *sha1, void *data)
+{
+       sha1_array_append(data, sha1);
+       return 0;
+}
+
 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
 {
+       struct sha1_array collect = SHA1_ARRAY_INIT;
        struct disambiguate_state ds;
+       int ret;
 
        if (init_object_disambiguation(prefix, strlen(prefix), &ds) < 0)
                return -1;
 
        ds.always_call_fn = 1;
-       ds.cb_data = cb_data;
-       ds.fn = fn;
-
+       ds.fn = collect_ambiguous;
+       ds.cb_data = &collect;
        find_short_object_filename(&ds);
        find_short_packed_object(&ds);
-       return ds.ambiguous;
+
+       ret = sha1_array_for_each_unique(&collect, fn, cb_data);
+       sha1_array_clear(&collect);
+       return ret;
+}
+
+/*
+ * Return the slot of the most-significant bit set in "val". There are various
+ * ways to do this quickly with fls() or __builtin_clzl(), but speed is
+ * probably not a big deal here.
+ */
+static unsigned msb(unsigned long val)
+{
+       unsigned r = 0;
+       while (val >>= 1)
+               r++;
+       return r;
 }
 
 int find_unique_abbrev_r(char *hex, const unsigned char *sha1, int len)
 {
        int status, exists;
 
+       if (len < 0) {
+               unsigned long count = approximate_object_count();
+               /*
+                * Add one because the MSB only tells us the highest bit set,
+                * not including the value of all the _other_ bits (so "15"
+                * is only one off of 2^4, but the MSB is the 3rd bit.
+                */
+               len = msb(count) + 1;
+               /*
+                * We now know we have on the order of 2^len objects, which
+                * expects a collision at 2^(len/2). But we also care about hex
+                * chars, not bits, and there are 4 bits per hex. So all
+                * together we need to divide by 2; but we also want to round
+                * odd numbers up, hence adding one before dividing.
+                */
+               len = (len + 1) / 2;
+               /*
+                * For very small repos, we stick with our regular fallback.
+                */
+               if (len < FALLBACK_DEFAULT_ABBREV)
+                       len = FALLBACK_DEFAULT_ABBREV;
+       }
+
        sha1_to_hex_r(hex, sha1);
        if (len == 40 || !len)
                return 40;
@@ -390,7 +508,10 @@ int find_unique_abbrev_r(char *hex, const unsigned char *sha1, int len)
 
 const char *find_unique_abbrev(const unsigned char *sha1, int len)
 {
-       static char hex[GIT_SHA1_HEXSZ + 1];
+       static int bufno;
+       static char hexbuffer[4][GIT_SHA1_HEXSZ + 1];
+       char *hex = hexbuffer[bufno];
+       bufno = (bufno + 1) % ARRAY_SIZE(hexbuffer);
        find_unique_abbrev_r(hex, sha1, len);
        return hex;
 }
@@ -930,7 +1051,7 @@ struct grab_nth_branch_switch_cbdata {
        struct strbuf buf;
 };
 
-static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
+static int grab_nth_branch_switch(struct object_id *ooid, struct object_id *noid,
                                  const char *email, unsigned long timestamp, int tz,
                                  const char *message, void *cb_data)
 {
@@ -1055,7 +1176,8 @@ static int interpret_empty_at(const char *name, int namelen, int len, struct str
        return 1;
 }
 
-static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
+static int reinterpret(const char *name, int namelen, int len,
+                      struct strbuf *buf, unsigned allowed)
 {
        /* we have extra data, which might need further processing */
        struct strbuf tmp = STRBUF_INIT;
@@ -1063,7 +1185,7 @@ static int reinterpret(const char *name, int namelen, int len, struct strbuf *bu
        int ret;
 
        strbuf_add(buf, name + len, namelen - len);
-       ret = interpret_branch_name(buf->buf, buf->len, &tmp);
+       ret = interpret_branch_name(buf->buf, buf->len, &tmp, allowed);
        /* that data was not interpreted, remove our cruft */
        if (ret < 0) {
                strbuf_setlen(buf, used);
@@ -1084,11 +1206,27 @@ static void set_shortened_ref(struct strbuf *buf, const char *ref)
        free(s);
 }
 
+static int branch_interpret_allowed(const char *refname, unsigned allowed)
+{
+       if (!allowed)
+               return 1;
+
+       if ((allowed & INTERPRET_BRANCH_LOCAL) &&
+           starts_with(refname, "refs/heads/"))
+               return 1;
+       if ((allowed & INTERPRET_BRANCH_REMOTE) &&
+           starts_with(refname, "refs/remotes/"))
+               return 1;
+
+       return 0;
+}
+
 static int interpret_branch_mark(const char *name, int namelen,
                                 int at, struct strbuf *buf,
                                 int (*get_mark)(const char *, int),
                                 const char *(*get_data)(struct branch *,
-                                                        struct strbuf *))
+                                                        struct strbuf *),
+                                unsigned allowed)
 {
        int len;
        struct branch *branch;
@@ -1113,64 +1251,55 @@ static int interpret_branch_mark(const char *name, int namelen,
        if (!value)
                die("%s", err.buf);
 
+       if (!branch_interpret_allowed(value, allowed))
+               return -1;
+
        set_shortened_ref(buf, value);
        return len + at;
 }
 
-/*
- * This reads short-hand syntax that not only evaluates to a commit
- * object name, but also can act as if the end user spelled the name
- * of the branch from the command line.
- *
- * - "@{-N}" finds the name of the Nth previous branch we were on, and
- *   places the name of the branch in the given buf and returns the
- *   number of characters parsed if successful.
- *
- * - "<branch>@{upstream}" finds the name of the other ref that
- *   <branch> is configured to merge with (missing <branch> defaults
- *   to the current branch), and places the name of the branch in the
- *   given buf and returns the number of characters parsed if
- *   successful.
- *
- * If the input is not of the accepted format, it returns a negative
- * number to signal an error.
- *
- * If the input was ok but there are not N branch switches in the
- * reflog, it returns 0.
- */
-int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
+int interpret_branch_name(const char *name, int namelen, struct strbuf *buf,
+                         unsigned allowed)
 {
        char *at;
        const char *start;
-       int len = interpret_nth_prior_checkout(name, namelen, buf);
+       int len;
 
        if (!namelen)
                namelen = strlen(name);
 
-       if (!len) {
-               return len; /* syntax Ok, not enough switches */
-       } else if (len > 0) {
-               if (len == namelen)
-                       return len; /* consumed all */
-               else
-                       return reinterpret(name, namelen, len, buf);
+       if (!allowed || (allowed & INTERPRET_BRANCH_LOCAL)) {
+               len = interpret_nth_prior_checkout(name, namelen, buf);
+               if (!len) {
+                       return len; /* syntax Ok, not enough switches */
+               } else if (len > 0) {
+                       if (len == namelen)
+                               return len; /* consumed all */
+                       else
+                               return reinterpret(name, namelen, len, buf, allowed);
+               }
        }
 
        for (start = name;
             (at = memchr(start, '@', namelen - (start - name)));
             start = at + 1) {
 
-               len = interpret_empty_at(name, namelen, at - name, buf);
-               if (len > 0)
-                       return reinterpret(name, namelen, len, buf);
+               if (!allowed || (allowed & INTERPRET_BRANCH_HEAD)) {
+                       len = interpret_empty_at(name, namelen, at - name, buf);
+                       if (len > 0)
+                               return reinterpret(name, namelen, len, buf,
+                                                  allowed);
+               }
 
                len = interpret_branch_mark(name, namelen, at - name, buf,
-                                           upstream_mark, branch_get_upstream);
+                                           upstream_mark, branch_get_upstream,
+                                           allowed);
                if (len > 0)
                        return len;
 
                len = interpret_branch_mark(name, namelen, at - name, buf,
-                                           push_mark, branch_get_push);
+                                           push_mark, branch_get_push,
+                                           allowed);
                if (len > 0)
                        return len;
        }
@@ -1178,22 +1307,19 @@ int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
        return -1;
 }
 
-int strbuf_branchname(struct strbuf *sb, const char *name)
+void strbuf_branchname(struct strbuf *sb, const char *name, unsigned allowed)
 {
        int len = strlen(name);
-       int used = interpret_branch_name(name, len, sb);
+       int used = interpret_branch_name(name, len, sb, allowed);
 
-       if (used == len)
-               return 0;
        if (used < 0)
                used = 0;
        strbuf_add(sb, name + used, len - used);
-       return len;
 }
 
 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
 {
-       strbuf_branchname(sb, name);
+       strbuf_branchname(sb, name, INTERPRET_BRANCH_LOCAL);
        if (name[0] == '-')
                return -1;
        strbuf_splice(sb, 0, 0, "refs/heads/", 11);