summaryrefslogtreecommitdiffstats
path: root/src
AgeCommit message (Collapse)AuthorFilesLines
2019-11-07xplist: Increase precision when converting PLIST_REAL nodes to XMLGravatar Nikias Bassen1-37/+16
2019-08-13plist: Remove unnecessary parameter from plist_copy_node()Gravatar Xiao Deng1-14/+6
2019-08-09Make sure to copy hash table entries properly when cloning array/dict nodesGravatar Nikias Bassen1-18/+22
As mentioned in #142, plist_copy_node() was not correctly handling the hash tables when cloning array or dict nodes; it incorrectly filled the hash table with the original child node info, which effectively would lead to a segmentation fault / UaF if the original array/dict would be freed followed by an attempt to access an element in the new hash table.
2019-07-07Add PList::Dictionary::const_iteratorGravatar Alexander Böhn1-0/+15
... and const versions of three member functions, each returning const_iterator: * Plist::Dictionary::Begin() * PList::Dictionary::End() * PList::Dictionary::Find()
2019-06-26Fixed bug in dictionary_fillGravatar tihmstar1-1/+1
Bug: when creating a new Dictionary object (for example through PList::Node::FromPlist(plist_t node) ), the dictionary_fill function is called from Dictionary() constructor in line 50. It seems that the intended way of calling dictionary_fill() is to pass the _map object by reference, however it is actually passed by value. Thus the changes to the map object made by dictionary_fill() are discarded when the function returns. Fix: pass _map by reference to keep the changes
2019-05-20plist_set_key_val(): prevent setting a key value that already exists in a ↵Gravatar Nikias Bassen1-0/+5
PLIST_DICT
2019-05-20Add plist_dict_item_get_key() to allow retrieving the key node for a given ↵Gravatar Nikias Bassen1-0/+11
item of a #PLIST_DICT
2019-05-19Add plist_array_item_remove() to allow removing an array's child node ↵Gravatar Nikias Bassen1-0/+15
without relying on the index
2019-05-19plist_array_get_item_index(): return UINT_MAX instead of 0 when node can't ↵Gravatar Nikias Bassen2-1/+5
be found
2019-05-19Add index lookup table for large PLIST_ARRAY nodesGravatar Nikias Bassen3-22/+113
2019-05-16Ignore invalid input in plist_get_*_val() to prevent unnecessary assertionsGravatar Nikias Bassen1-19/+50
Also fixes #126 by skipping the strlen() in the assert() if for some reason NULL is returned as data
2019-05-11Use Requires.private in *.pc files to not add unnecessary links.Gravatar Julien Lavergne1-1/+1
2019-01-21cpp: Dictionary: Reduce code duplication with helper functionGravatar Nikias Bassen1-47/+16
2019-01-21plist: Add iterator for #PLIST_ARRAY nodesGravatar Nikias Bassen2-26/+50
Similar to #PLIST_DICT, an iterator can now be used for #PLIST_ARRAY nodes. Get an iterator with plist_array_new_iter() and use plist_array_next_item() to iterate over the elements.
2019-01-21xplist: Fix dtostr() inserting ':' to output string when fractional part ↵Gravatar Nikias Bassen1-2/+15
overflows
2019-01-18bplist: Silence compiler warning 'comparison is always true ...' (32 bit)Gravatar Nikias Bassen1-1/+1
2018-12-23plist: Improve plist_dict_next_item() drastically by iterating on node list ↵Gravatar Nikias Bassen1-10/+8
directly As Xiao Deng pointed out in #131, plist_dict_next_item() was very inefficient. For each iteration, node_nth_child() was called with the iterator value, which would walk through the child node list on EVERY iteration. If the dictionary is large this makes things very slow. More than that, after reaching the key node the code was calling node_nth_child() AGAIN (with iterator value + 1) to reach the value node, which would walk through the node list once more. This commit changes the iterator to be a node_t pointer so that the iteration is done on the node list directly.
2018-12-17xplist: Write base64 directly to output buffer to improve memory usageGravatar Nikias Bassen1-5/+2
Now that we grow the output buffer enough before writing XML output we can just write the base64 encoded data directly to the ouput buffer instead of using a heap buffer that will then be copied to the output buffer. This makes writing XML output more memory efficient (and slightly faster).
2018-12-14xplist: Improve memory usage by estimating output buffer sizeGravatar Nikias Bassen2-3/+128
2018-12-14xplist: Fix writing of empty dict and array nodes to XMLGravatar Nikias Bassen1-8/+6
2018-12-11xplist: Fix segfault caused by recent changes in libcnaryGravatar Nikias Bassen1-1/+1
2018-12-10ptrarray: Allow larger chunks for buffer reallocationGravatar Nikias Bassen1-1/+1
2018-12-10xplist: Prevent unnecessary reallocations when writing XML outputGravatar Nikias Bassen1-1/+4
2018-12-10Remove node_iterator and operate on node list directly to improve memory usageGravatar Nikias Bassen3-15/+6
2018-12-10bplist: Improve performance and memory usage when writing binary plistGravatar Nikias Bassen4-6/+88
2018-11-30bplist: Remove unnecessary allocations when parsing and writing unicode nodesGravatar Nikias Bassen1-88/+72
2018-11-29bplist: Remove redundant calls to strlen()Gravatar Nikias Bassen1-7/+4
2018-09-04xplist: Assert when number of child nodes of PLIST_DICT is not evenGravatar Nikias Bassen1-0/+3
This should only happen due to misuse of the library, e.g. when calling plist_free() on a node that is a value node in a PLIST_DICT without properly removing the dictionary entry (key/value pair) and then calling plist_to_xml() on that dictionary.
2018-07-25xplist: Fix typo in error messageGravatar Bastien Nocera1-1/+1
2017-05-31bplist: Prevent store to misaligned address when writing real/date nodesGravatar Nikias Bassen1-9/+9
ASAN reported possible undefined behaviour when writing float/double values to misaligned addresses.
2017-05-29bplist: Work around misaligned reads reported by AddressSanitizerGravatar Nikias Bassen1-3/+3
These misaligned reads reported by ASAN might lead to undefined behavior.
2017-04-20bplist: Fix missing break in switch statement in plist_to_bin()Gravatar Nikias Bassen1-0/+1
Credit to Christophe Fergeau
2017-04-20bplist: Suppress compiler warnings with proper castsGravatar Nikias Bassen1-3/+3
2017-04-19bplist: Fix integer overflow check (offset table size)Gravatar Nikias Bassen1-3/+17
2017-04-14Initialize safe_year in time64.cGravatar Greg Dennis1-2/+2
Clang fails with stricter compilation options, because it thinks safe_year may be uninitialized at the return statement. The logic prevents it from being uninitialized, but probably worth the initialization to avoid the compiler error. The rest of libimobiledevice compiles successfully under the same options.
2017-04-06Update time64_limits.hGravatar Greg Dennis1-0/+2
This depends on the 'tm' type being declared, which is defined in time.h.
2017-04-02xplist: Plug another memory leakGravatar Nikias Bassen1-0/+3
Credit to OSS-Fuzz
2017-03-29xplist: Prevent memory leak(s) when parsing failsGravatar Nikias Bassen1-2/+2
Credit to OSS-Fuzz
2017-03-29xplist: Make XML parsing non-recursive to prevent stack overflow on ↵Gravatar Nikias Bassen1-79/+103
deep-structured plists Credit to OSS-Fuzz
2017-03-26bplist: Make sure sanity checks work on 32bit platformsGravatar Nikias Bassen1-10/+14
Because on 32-bit platforms 32-bit pointers and 64-bit sizes have been used for the sanity checks of the offset table and object references, the range checks would fail in certain interger-overflowish situations, causing heap buffer overflows or other unwanted behavior. Fixed by wideing the operands in question to 64-bit.
2017-02-18base64: Prevent undefined shift when parsing invalid base64 encoded dataGravatar Nikias Bassen1-3/+3
Credit to OSS-Fuzz
2017-02-15xplist: Improve writing of large PLIST_DATA nodes by growing buffer in advanceGravatar Nikias Bassen4-3/+11
Instead of letting the buffer grow by just the amount of bytes currently transformed to base64 - which is basically line by line - we now calculate the size of the output blob in advance and grow the buffer accordingly. This will reduce the amount of reallocs to just one, which is especially important for large data blobs. While this is a general improvement for all platforms, it is on platforms like Windows where realloc() can be REALLY slow; converting a 20mb blob to XML can easily take up to a minute (due to the several hundred thousand calls to realloc()). With this commit, it will be fast again.
2017-02-10bplist: Fix data range check for string/data/dict/array nodesGravatar Nikias Bassen1-6/+6
Passing a size of 0xFFFFFFFFFFFFFFFF to parse_string_node() might result in a memcpy with a size of -1, leading to undefined behavior. This commit makes sure that the actual node data (which depends on the size) is in the range start_of_object..start_of_object+size. Credit to OSS-Fuzz
2017-02-10bplist: Fix integer overflow resulting in OOB heap buffer readGravatar Nikias Bassen1-0/+5
Credit to OSS-Fuzz
2017-02-09xplist: Fix OOB heap buffer read with empty data nodesGravatar Nikias Bassen1-2/+4
Credit to OSS-Fuzz
2017-02-09bplist: Make sure to detect integer overflow when handling unicode node sizeGravatar Nikias Bassen1-0/+4
Credit to OSS-Fuzz
2017-02-09xplist: Prevent assert when parsing CF$UID dict with invalid value nodeGravatar Nikias Bassen1-0/+5
Credit to OSS-Fuzz
2017-02-08xplist: Use proper variable size for integer from string parsingGravatar Nikias Bassen1-2/+2
2017-02-07plist: Fix assert() to allow 16 or 8 byte integer sizes (16 bytes = unsigned ↵Gravatar Nikias Bassen1-1/+1
integer) Credit to Wang Junjie <zhunkibatu@gmail.com> (#90) Credit to OSS-Fuzz
2017-02-07bplist: Properly handle some more malloc() failure situationsGravatar Nikias Bassen1-3/+18