From: Nguyen Thai Ngoc Duy Date: Tue, 07 Sep 2010 01:28:10 GMT Subject: Re: [RFC PATCH 07/15] cache_tree_update(): Capability to handle tree entries missing from index Message-ID: In-Reply-To: <1283645647-1891-8-git-send-email-newren@gmail.com> On Sun, Sep 5, 2010 at 10:13 AM, Elijah Newren wrote: > +static void add_missing_paths_before(const char *path, > +                                    int pathlen, > +                                    int baselen, > +                                    struct tree_desc *desc, > +                                    struct strbuf *buffer) > +{ > +       if (!git_sparse_pathspecs) > +               return; /* No paths are missing */ > + > +       for (; desc->size; update_tree_entry(desc)) { > +               struct name_entry entry = desc->entry; > +               int entlen = strlen(entry.path); > + > +               /* We only want paths before path */ > +               if (path) > +                       /* > +                        * FIXME: if entlen < pathlen, do I need to > +                        * use strncmp instead? Does entry.path being > +                        * NUL-terminated ensure memcmp exits > +                        * early? > +                        */ > +                       if (memcmp(entry.path, > +                                  path+baselen, > +                                  pathlen-baselen) >= 0) > +                               break; If you do "break;" here, the current entry remains in "desc" and will be used to regenerate a duplicate tree in add_missing_paths_before(NULL, 0, 0, ..);. That entry is added the first time by update_one itself before the second add_missing() call. This fixes it for me diff --git a/cache-tree.c b/cache-tree.c index 79c28cc..1655738 100644 --- a/cache-tree.c +++ b/cache-tree.c @@ -278,17 +278,19 @@ static void add_missing_paths_before(const char *path, int entlen = strlen(entry.path); /* We only want paths before path */ - if (path) + if (path) { /* * FIXME: if entlen < pathlen, do I need to * use strncmp instead? Does entry.path being * NUL-terminated ensure memcmp exits * early? */ - if (memcmp(entry.path, - path+baselen, - pathlen-baselen) >= 0) + int result = memcmp(entry.path, path+baselen, pathlen-baselen); + if (result > 0) break; + if (!result) + continue; + } /* We only want paths "missing" from index due to sparsity */ if (path) { -- Duy