X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fdentry.c;h=d93d8ac739af64a8254b994fc078b227aaac14d4;hp=b7a0032aa7be7e9bdcf7827a5e17ba06b22ca028;hb=45663beea6d4d50f32bf6f8f0865e6d26a27537b;hpb=1c940d499beb539f846ed2509e3fee7772adf250 diff --git a/src/dentry.c b/src/dentry.c index b7a0032a..d93d8ac7 100644 --- a/src/dentry.c +++ b/src/dentry.c @@ -70,6 +70,8 @@ struct wim_ads_entry_on_disk { utf16lechar stream_name[]; } _packed_attribute; +#define WIM_ADS_ENTRY_DISK_SIZE 38 + /* WIM directory entry (on-disk format) */ struct wim_dentry_on_disk { le64 length; @@ -105,6 +107,8 @@ struct wim_dentry_on_disk { /*utf16lechar short_name[];*/ } _packed_attribute; +#define WIM_DENTRY_DISK_SIZE 102 + /* Calculates the unaligned length, in bytes, of an on-disk WIM dentry that has * a file name and short name that take the specified numbers of bytes. This * excludes any alternate data stream entries that may follow the dentry. */ @@ -130,16 +134,6 @@ dentry_correct_length_unaligned(const struct wim_dentry *dentry) dentry->short_name_nbytes); } -/* Return %true iff the alternate data stream entry @entry has the UTF-16LE - * stream name @name that has length @name_nbytes bytes. */ -static inline bool -ads_entry_has_name(const struct wim_ads_entry *entry, - const utf16lechar *name, size_t name_nbytes) -{ - return entry->stream_name_nbytes == name_nbytes && - memcmp(entry->stream_name, name, name_nbytes) == 0; -} - /* Duplicates a string of system-dependent encoding into a UTF-16LE string and * returns the string and its length, in bytes, in the pointer arguments. Frees * any existing string at the return location before overwriting it. */ @@ -485,25 +479,100 @@ calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p) } } +/* UNIX: Case-sensitive UTF-16LE dentry or stream name comparison. We call this + * on Windows as well to distinguish true duplicates from names differing by + * case only. */ static int -compare_utf16le_names(const utf16lechar *name1, size_t nbytes1, - const utf16lechar *name2, size_t nbytes2) +compare_utf16le_names_case_sensitive(const utf16lechar *name1, size_t nbytes1, + const utf16lechar *name2, size_t nbytes2) { + /* Return the result if the strings differ up to their minimum length. + * Note that we cannot use strcmp() or strncmp() here, as the strings + * are in UTF-16LE format. */ int result = memcmp(name1, name2, min(nbytes1, nbytes2)); if (result) return result; + + /* The strings are the same up to their minimum length, so return a + * result based on their lengths. */ + if (nbytes1 < nbytes2) + return -1; + else if (nbytes1 > nbytes2) + return 1; else - return (int)nbytes1 - (int)nbytes2; + return 0; } +#ifdef __WIN32__ +/* Windoze: Case-insensitive UTF-16LE dentry or stream name comparison */ static int -dentry_compare_names(const struct wim_dentry *d1, const struct wim_dentry *d2) +compare_utf16le_names_case_insensitive(const utf16lechar *name1, size_t nbytes1, + const utf16lechar *name2, size_t nbytes2) { - return compare_utf16le_names(d1->file_name, d1->file_name_nbytes, - d2->file_name, d2->file_name_nbytes); + /* Return the result if the strings differ up to their minimum length. + * */ + int result = _wcsnicmp((const wchar_t*)name1, (const wchar_t*)name2, + min(nbytes1 / 2, nbytes2 / 2)); + if (result) + return result; + + /* The strings are the same up to their minimum length, so return a + * result based on their lengths. */ + if (nbytes1 < nbytes2) + return -1; + else if (nbytes1 > nbytes2) + return 1; + else + return 0; } +#endif /* __WIN32__ */ + +#ifdef __WIN32__ +# define compare_utf16le_names compare_utf16le_names_case_insensitive +#else +# define compare_utf16le_names compare_utf16le_names_case_sensitive +#endif +#ifdef __WIN32__ +static int +dentry_compare_names_case_insensitive(const struct wim_dentry *d1, + const struct wim_dentry *d2) +{ + return compare_utf16le_names_case_insensitive(d1->file_name, + d1->file_name_nbytes, + d2->file_name, + d2->file_name_nbytes); +} +#endif /* __WIN32__ */ + +static int +dentry_compare_names_case_sensitive(const struct wim_dentry *d1, + const struct wim_dentry *d2) +{ + return compare_utf16le_names_case_sensitive(d1->file_name, + d1->file_name_nbytes, + d2->file_name, + d2->file_name_nbytes); +} + +#ifdef __WIN32__ +# define dentry_compare_names dentry_compare_names_case_insensitive +#else +# define dentry_compare_names dentry_compare_names_case_sensitive +#endif + +/* Return %true iff the alternate data stream entry @entry has the UTF-16LE + * stream name @name that has length @name_nbytes bytes. */ +static inline bool +ads_entry_has_name(const struct wim_ads_entry *entry, + const utf16lechar *name, size_t name_nbytes) +{ + return !compare_utf16le_names(name, name_nbytes, + entry->stream_name, + entry->stream_name_nbytes); +} + struct wim_dentry * get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry, const utf16lechar *name, @@ -1018,7 +1087,9 @@ free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table) * @parent: The dentry that will be the parent of @child. * @child: The dentry to link. * - * Returns non-NULL if a duplicate dentry was detected. + * Returns NULL if successful. If @parent already contains a dentry with the + * same name as @child (see compare_utf16le_names() for what names are + * considered the "same"), the pointer to this duplicate dentry is returned. */ struct wim_dentry * dentry_add_child(struct wim_dentry * restrict parent, @@ -1180,12 +1251,11 @@ add_stream_from_data_buffer(const void *buffer, size_t size, lte = new_lookup_table_entry(); if (!lte) return NULL; - buffer_copy = MALLOC(size); + buffer_copy = memdup(buffer, size); if (!buffer_copy) { free_lookup_table_entry(lte); return NULL; } - memcpy(buffer_copy, buffer, size); lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER; lte->attached_buffer = buffer_copy; lte->resource_entry.original_size = size; @@ -1388,6 +1458,8 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode, struct wim_ads_entry *ads_entries; int ret; + BUILD_BUG_ON(sizeof(struct wim_ads_entry_on_disk) != WIM_ADS_ENTRY_DISK_SIZE); + /* Allocate an array for our in-memory representation of the alternate * data stream entries. */ num_ads = inode->i_num_ads; @@ -1532,6 +1604,8 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, const struct wim_dentry_on_disk *disk_dentry = (const struct wim_dentry_on_disk*)p; + BUILD_BUG_ON(sizeof(struct wim_dentry_on_disk) != WIM_DENTRY_DISK_SIZE); + if ((uintptr_t)p & 7) WARNING("WIM dentry is not 8-byte aligned"); @@ -1725,6 +1799,18 @@ out: return ret; } +static const tchar * +dentry_get_file_type_string(const struct wim_dentry *dentry) +{ + const struct wim_inode *inode = dentry->d_inode; + if (inode_is_directory(inode)) + return T("directory"); + else if (inode_is_symlink(inode)) + return T("symbolic link"); + else + return T("file"); +} + /* Reads the children of a dentry, and all their children, ..., etc. from the * metadata resource and into the dentry tree. * @@ -1747,6 +1833,7 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, { u64 cur_offset = dentry->subdir_offset; struct wim_dentry *child; + struct wim_dentry *duplicate; struct wim_dentry cur_child; int ret; @@ -1774,14 +1861,12 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, /* Not end of directory. Allocate this child permanently and * link it to the parent and previous child. */ - child = MALLOC(sizeof(struct wim_dentry)); + child = memdup(&cur_child, sizeof(struct wim_dentry)); if (!child) { - ERROR("Failed to allocate %zu bytes for new dentry", - sizeof(struct wim_dentry)); + ERROR("Failed to allocate new dentry!"); ret = WIMLIB_ERR_NOMEM; break; } - memcpy(child, &cur_child, sizeof(struct wim_dentry)); /* Advance to the offset of the next child. Note: We need to * advance by the TOTAL length of the dentry, not by the length @@ -1790,17 +1875,41 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, * entries. */ cur_offset += dentry_total_length(child); - if (dentry_add_child(dentry, child)) { - WARNING("Ignoring duplicate dentry \"%"WS"\"", - child->file_name); - WARNING("(In directory \"%"TS"\")", dentry_full_path(dentry)); + duplicate = dentry_add_child(dentry, child); + if (duplicate) { + const tchar *child_type, *duplicate_type; + child_type = dentry_get_file_type_string(child); + duplicate_type = dentry_get_file_type_string(duplicate); + /* On UNIX, duplicates are exact. On Windows, + * duplicates may differ by case and we wish to provide + * a different warning message in this case. */ + #ifdef __WIN32__ + if (dentry_compare_names_case_sensitive(child, duplicate)) + { + child->parent = dentry; + WARNING("Ignoring %ls \"%ls\", which differs " + "only in case from %ls \"%ls\"", + child_type, + dentry_full_path(child), + duplicate_type, + dentry_full_path(duplicate)); + } + else + #endif + { + WARNING("Ignoring duplicate %"TS" \"%"TS"\" " + "(the WIM image already contains a %"TS" " + "at that path with the exact same name)", + child_type, dentry_full_path(duplicate), + duplicate_type); + } free_dentry(child); } else { inode_add_dentry(child, child->d_inode); /* If there are children of this child, call this * procedure recursively. */ if (child->subdir_offset != 0) { - if (!dentry_is_directory(child)) { + if (dentry_is_directory(child)) { ret = read_dentry_tree(metadata_resource, metadata_resource_len, child);