git-tar-tree: no more void pointer arithmetic
[git.git] / http-fetch.c
index 555e95d..da1a7f5 100644 (file)
@@ -4,16 +4,47 @@
 #include "fetch.h"
 #include "http.h"
 
+#ifndef NO_EXPAT
+#include <expat.h>
+
+/* Definitions for DAV requests */
+#define DAV_PROPFIND "PROPFIND"
+#define DAV_PROPFIND_RESP ".multistatus.response"
+#define DAV_PROPFIND_NAME ".multistatus.response.href"
+#define DAV_PROPFIND_COLLECTION ".multistatus.response.propstat.prop.resourcetype.collection"
+#define PROPFIND_ALL_REQUEST "<?xml version=\"1.0\" encoding=\"utf-8\" ?>\n<D:propfind xmlns:D=\"DAV:\">\n<D:allprop/>\n</D:propfind>"
+
+/* Definitions for processing XML DAV responses */
+#ifndef XML_STATUS_OK
+enum XML_Status {
+  XML_STATUS_OK = 1,
+  XML_STATUS_ERROR = 0
+};
+#define XML_STATUS_OK    1
+#define XML_STATUS_ERROR 0
+#endif
+
+/* Flags that control remote_ls processing */
+#define PROCESS_FILES (1u << 0)
+#define PROCESS_DIRS  (1u << 1)
+#define RECURSIVE     (1u << 2)
+
+/* Flags that remote_ls passes to callback functions */
+#define IS_DIR (1u << 0)
+#endif
+
 #define PREV_BUF_SIZE 4096
 #define RANGE_HEADER_SIZE 30
 
 static int got_alternates = -1;
+static int corrupt_object_found = 0;
 
 static struct curl_slist *no_pragma_header;
 
 struct alt_base
 {
        char *base;
+       int path_len;
        int got_indices;
        struct packed_git *packs;
        struct alt_base *next;
@@ -57,6 +88,30 @@ struct alternates_request {
        int http_specific;
 };
 
+#ifndef NO_EXPAT
+struct xml_ctx
+{
+       char *name;
+       int len;
+       char *cdata;
+       void (*userFunc)(struct xml_ctx *ctx, int tag_closed);
+       void *userData;
+};
+
+struct remote_ls_ctx
+{
+       struct alt_base *repo;
+       char *path;
+       void (*userFunc)(struct remote_ls_ctx *ls);
+       void *userData;
+       int flags;
+       char *dentry_name;
+       int dentry_flags;
+       int rc;
+       struct remote_ls_ctx *parent;
+};
+#endif
+
 static struct object_request *object_queue_head = NULL;
 
 static size_t fwrite_sha1_file(void *ptr, size_t eltsize, size_t nmemb,
@@ -130,7 +185,7 @@ static void start_object_request(struct object_request *obj_req)
 
        if (obj_req->local < 0) {
                obj_req->state = ABORTED;
-               error("Couldn't create temporary file %s for %s: %s\n",
+               error("Couldn't create temporary file %s for %s: %s",
                      obj_req->tmpfile, obj_req->filename, strerror(errno));
                return;
        }
@@ -344,6 +399,7 @@ void prefetch(unsigned char *sha1)
        snprintf(newreq->filename, sizeof(newreq->filename), "%s", filename);
        snprintf(newreq->tmpfile, sizeof(newreq->tmpfile),
                 "%s.temp", filename);
+       newreq->slot = NULL;
        newreq->next = NULL;
 
        if (object_queue_head == NULL) {
@@ -468,13 +524,15 @@ static void process_alternates_response(void *callback_data)
                                         alt_req->url);
                        active_requests++;
                        slot->in_use = 1;
-                       if (start_active_slot(slot)) {
-                               return;
-                       } else {
+                       if (slot->finished != NULL)
+                               (*slot->finished) = 0;
+                       if (!start_active_slot(slot)) {
                                got_alternates = -1;
                                slot->in_use = 0;
-                               return;
+                               if (slot->finished != NULL)
+                                       (*slot->finished) = 1;
                        }
+                       return;
                }
        } else if (slot->curl_result != CURLE_OK) {
                if (slot->http_code != 404 &&
@@ -497,6 +555,7 @@ static void process_alternates_response(void *callback_data)
                        int serverlen = 0;
                        struct alt_base *newalt;
                        char *target = NULL;
+                       char *path;
                        if (data[i] == '/') {
                                serverlen = strchr(base + 8, '/') - base;
                                okay = 1;
@@ -525,10 +584,8 @@ static void process_alternates_response(void *callback_data)
                        // skip 'objects' at end
                        if (okay) {
                                target = xmalloc(serverlen + posn - i - 6);
-                               strncpy(target, base, serverlen);
-                               strncpy(target + serverlen, data + i,
-                                       posn - i - 7);
-                               target[serverlen + posn - i - 7] = '\0';
+                               safe_strncpy(target, base, serverlen);
+                               safe_strncpy(target + serverlen, data + i, posn - i - 6);
                                if (get_verbosely)
                                        fprintf(stderr,
                                                "Also look at %s\n", target);
@@ -537,6 +594,13 @@ static void process_alternates_response(void *callback_data)
                                newalt->base = target;
                                newalt->got_indices = 0;
                                newalt->packs = NULL;
+                               path = strstr(target, "//");
+                               if (path) {
+                                       path = strchr(path+2, '/');
+                                       if (path)
+                                               newalt->path_len = strlen(path);
+                               }
+
                                while (tail->next != NULL)
                                        tail = tail->next;
                                tail->next = newalt;
@@ -608,6 +672,209 @@ static void fetch_alternates(char *base)
        free(url);
 }
 
+#ifndef NO_EXPAT
+static void
+xml_start_tag(void *userData, const char *name, const char **atts)
+{
+       struct xml_ctx *ctx = (struct xml_ctx *)userData;
+       const char *c = strchr(name, ':');
+       int new_len;
+
+       if (c == NULL)
+               c = name;
+       else
+               c++;
+
+       new_len = strlen(ctx->name) + strlen(c) + 2;
+
+       if (new_len > ctx->len) {
+               ctx->name = xrealloc(ctx->name, new_len);
+               ctx->len = new_len;
+       }
+       strcat(ctx->name, ".");
+       strcat(ctx->name, c);
+
+       if (ctx->cdata) {
+               free(ctx->cdata);
+               ctx->cdata = NULL;
+       }
+
+       ctx->userFunc(ctx, 0);
+}
+
+static void
+xml_end_tag(void *userData, const char *name)
+{
+       struct xml_ctx *ctx = (struct xml_ctx *)userData;
+       const char *c = strchr(name, ':');
+       char *ep;
+
+       ctx->userFunc(ctx, 1);
+
+       if (c == NULL)
+               c = name;
+       else
+               c++;
+
+       ep = ctx->name + strlen(ctx->name) - strlen(c) - 1;
+       *ep = 0;
+}
+
+static void
+xml_cdata(void *userData, const XML_Char *s, int len)
+{
+       struct xml_ctx *ctx = (struct xml_ctx *)userData;
+       if (ctx->cdata)
+               free(ctx->cdata);
+       ctx->cdata = xmalloc(len + 1);
+       safe_strncpy(ctx->cdata, s, len + 1);
+}
+
+static int remote_ls(struct alt_base *repo, const char *path, int flags,
+                    void (*userFunc)(struct remote_ls_ctx *ls),
+                    void *userData);
+
+static void handle_remote_ls_ctx(struct xml_ctx *ctx, int tag_closed)
+{
+       struct remote_ls_ctx *ls = (struct remote_ls_ctx *)ctx->userData;
+
+       if (tag_closed) {
+               if (!strcmp(ctx->name, DAV_PROPFIND_RESP) && ls->dentry_name) {
+                       if (ls->dentry_flags & IS_DIR) {
+                               if (ls->flags & PROCESS_DIRS) {
+                                       ls->userFunc(ls);
+                               }
+                               if (strcmp(ls->dentry_name, ls->path) &&
+                                   ls->flags & RECURSIVE) {
+                                       ls->rc = remote_ls(ls->repo,
+                                                          ls->dentry_name,
+                                                          ls->flags,
+                                                          ls->userFunc,
+                                                          ls->userData);
+                               }
+                       } else if (ls->flags & PROCESS_FILES) {
+                               ls->userFunc(ls);
+                       }
+               } else if (!strcmp(ctx->name, DAV_PROPFIND_NAME) && ctx->cdata) {
+                       ls->dentry_name = xmalloc(strlen(ctx->cdata) -
+                                                 ls->repo->path_len + 1);
+                       strcpy(ls->dentry_name, ctx->cdata + ls->repo->path_len);
+               } else if (!strcmp(ctx->name, DAV_PROPFIND_COLLECTION)) {
+                       ls->dentry_flags |= IS_DIR;
+               }
+       } else if (!strcmp(ctx->name, DAV_PROPFIND_RESP)) {
+               if (ls->dentry_name) {
+                       free(ls->dentry_name);
+               }
+               ls->dentry_name = NULL;
+               ls->dentry_flags = 0;
+       }
+}
+
+static int remote_ls(struct alt_base *repo, const char *path, int flags,
+                    void (*userFunc)(struct remote_ls_ctx *ls),
+                    void *userData)
+{
+       char *url = xmalloc(strlen(repo->base) + strlen(path) + 1);
+       struct active_request_slot *slot;
+       struct slot_results results;
+       struct buffer in_buffer;
+       struct buffer out_buffer;
+       char *in_data;
+       char *out_data;
+       XML_Parser parser = XML_ParserCreate(NULL);
+       enum XML_Status result;
+       struct curl_slist *dav_headers = NULL;
+       struct xml_ctx ctx;
+       struct remote_ls_ctx ls;
+
+       ls.flags = flags;
+       ls.repo = repo;
+       ls.path = strdup(path);
+       ls.dentry_name = NULL;
+       ls.dentry_flags = 0;
+       ls.userData = userData;
+       ls.userFunc = userFunc;
+       ls.rc = 0;
+
+       sprintf(url, "%s%s", repo->base, path);
+
+       out_buffer.size = strlen(PROPFIND_ALL_REQUEST);
+       out_data = xmalloc(out_buffer.size + 1);
+       snprintf(out_data, out_buffer.size + 1, PROPFIND_ALL_REQUEST);
+       out_buffer.posn = 0;
+       out_buffer.buffer = out_data;
+
+       in_buffer.size = 4096;
+       in_data = xmalloc(in_buffer.size);
+       in_buffer.posn = 0;
+       in_buffer.buffer = in_data;
+
+       dav_headers = curl_slist_append(dav_headers, "Depth: 1");
+       dav_headers = curl_slist_append(dav_headers, "Content-Type: text/xml");
+
+       slot = get_active_slot();
+       slot->results = &results;
+       curl_easy_setopt(slot->curl, CURLOPT_INFILE, &out_buffer);
+       curl_easy_setopt(slot->curl, CURLOPT_INFILESIZE, out_buffer.size);
+       curl_easy_setopt(slot->curl, CURLOPT_READFUNCTION, fread_buffer);
+       curl_easy_setopt(slot->curl, CURLOPT_FILE, &in_buffer);
+       curl_easy_setopt(slot->curl, CURLOPT_WRITEFUNCTION, fwrite_buffer);
+       curl_easy_setopt(slot->curl, CURLOPT_URL, url);
+       curl_easy_setopt(slot->curl, CURLOPT_UPLOAD, 1);
+       curl_easy_setopt(slot->curl, CURLOPT_CUSTOMREQUEST, DAV_PROPFIND);
+       curl_easy_setopt(slot->curl, CURLOPT_HTTPHEADER, dav_headers);
+
+       if (start_active_slot(slot)) {
+               run_active_slot(slot);
+               if (results.curl_result == CURLE_OK) {
+                       ctx.name = xcalloc(10, 1);
+                       ctx.len = 0;
+                       ctx.cdata = NULL;
+                       ctx.userFunc = handle_remote_ls_ctx;
+                       ctx.userData = &ls;
+                       XML_SetUserData(parser, &ctx);
+                       XML_SetElementHandler(parser, xml_start_tag,
+                                             xml_end_tag);
+                       XML_SetCharacterDataHandler(parser, xml_cdata);
+                       result = XML_Parse(parser, in_buffer.buffer,
+                                          in_buffer.posn, 1);
+                       free(ctx.name);
+
+                       if (result != XML_STATUS_OK) {
+                               ls.rc = error("XML error: %s",
+                                             XML_ErrorString(
+                                                     XML_GetErrorCode(parser)));
+                       }
+               } else {
+                       ls.rc = -1;
+               }
+       } else {
+               ls.rc = error("Unable to start PROPFIND request");
+       }
+
+       free(ls.path);
+       free(url);
+       free(out_data);
+       free(in_buffer.buffer);
+       curl_slist_free_all(dav_headers);
+
+       return ls.rc;
+}
+
+static void process_ls_pack(struct remote_ls_ctx *ls)
+{
+       unsigned char sha1[20];
+
+       if (strlen(ls->dentry_name) == 63 &&
+           !strncmp(ls->dentry_name, "objects/pack/pack-", 18) &&
+           !strncmp(ls->dentry_name+58, ".pack", 5)) {
+               get_sha1_hex(ls->dentry_name + 18, sha1);
+               setup_index(ls->repo, sha1);
+       }
+}
+#endif
+
 static int fetch_indices(struct alt_base *repo)
 {
        unsigned char sha1[20];
@@ -630,6 +897,12 @@ static int fetch_indices(struct alt_base *repo)
        if (get_verbosely)
                fprintf(stderr, "Getting pack list for %s\n", repo->base);
 
+#ifndef NO_EXPAT
+       if (remote_ls(repo, "objects/pack/", PROCESS_FILES,
+                     process_ls_pack, NULL) == 0)
+               return 0;
+#endif
+
        url = xmalloc(strlen(repo->base) + 21);
        sprintf(url, "%s/objects/info/packs", repo->base);
 
@@ -775,6 +1048,20 @@ static int fetch_pack(struct alt_base *repo, unsigned char *sha1)
        return 0;
 }
 
+static void abort_object_request(struct object_request *obj_req)
+{
+       if (obj_req->local >= 0) {
+               close(obj_req->local);
+               obj_req->local = -1;
+       }
+       unlink(obj_req->tmpfile);
+       if (obj_req->slot) {
+               release_active_slot(obj_req->slot);
+               obj_req->slot = NULL;
+       }
+       release_object_request(obj_req);
+}
+
 static int fetch_object(struct alt_base *repo, unsigned char *sha1)
 {
        char *hex = sha1_to_hex(sha1);
@@ -787,7 +1074,7 @@ static int fetch_object(struct alt_base *repo, unsigned char *sha1)
                return error("Couldn't find request for %s in the queue", hex);
 
        if (has_sha1_file(obj_req->sha1)) {
-               release_object_request(obj_req);
+               abort_object_request(obj_req);
                return 0;
        }
 
@@ -818,9 +1105,10 @@ static int fetch_object(struct alt_base *repo, unsigned char *sha1)
                                    obj_req->errorstr, obj_req->curl_result,
                                    obj_req->http_code, hex);
        } else if (obj_req->zret != Z_STREAM_END) {
-               ret = error("File %s (%s) corrupt\n", hex, obj_req->url);
+               corrupt_object_found++;
+               ret = error("File %s (%s) corrupt", hex, obj_req->url);
        } else if (memcmp(obj_req->sha1, obj_req->real_sha1, 20)) {
-               ret = error("File %s has bad hash\n", hex);
+               ret = error("File %s has bad hash", hex);
        } else if (obj_req->rename < 0) {
                ret = error("unable to write sha1 filename %s",
                            obj_req->filename);
@@ -842,7 +1130,7 @@ int fetch(unsigned char *sha1)
                fetch_alternates(alt->base);
                altbase = altbase->next;
        }
-       return error("Unable to find %s under %s\n", sha1_to_hex(sha1),
+       return error("Unable to find %s under %s", sha1_to_hex(sha1),
                     alt->base);
 }
 
@@ -929,10 +1217,12 @@ int main(int argc, char **argv)
 {
        char *commit_id;
        char *url;
+       char *path;
        int arg = 1;
        int rc = 0;
 
        setup_git_directory();
+       git_config(git_default_config);
 
        while (arg < argc && argv[arg][0] == '-') {
                if (argv[arg][1] == 't') {
@@ -959,6 +1249,7 @@ int main(int argc, char **argv)
        }
        commit_id = argv[arg];
        url = argv[arg + 1];
+       write_ref_log_details = url;
 
        http_init();
 
@@ -969,13 +1260,25 @@ int main(int argc, char **argv)
        alt->got_indices = 0;
        alt->packs = NULL;
        alt->next = NULL;
+       path = strstr(url, "//");
+       if (path) {
+               path = strchr(path+2, '/');
+               if (path)
+                       alt->path_len = strlen(path);
+       }
 
        if (pull(commit_id))
                rc = 1;
 
-       curl_slist_free_all(no_pragma_header);
-
        http_cleanup();
 
+       curl_slist_free_all(no_pragma_header);
+
+       if (corrupt_object_found) {
+               fprintf(stderr,
+"Some loose object were found to be corrupt, but they might be just\n"
+"a false '404 Not Found' error message sent with incorrect HTTP\n"
+"status code.  Suggest running git fsck-objects.\n");
+       }
        return rc;
 }