tfussell / miniz-cpp Goto Github PK

I ran Valgrind on my program, that at some point uses miniz-cpp.

It raised one error for miniz-cpp, which is the following:

==37967== Conditional jump or move depends on uninitialised value(s)
==37967==    at 0x565D9C4: tdefl_find_match (zip_file.hpp:2339)
==37967==    by 0x565D9C4: tdefl_compress_normal (zip_file.hpp:2610)
==37967==    by 0x565EA34: tdefl_compress (zip_file.hpp:2727)
==37967==    by 0x565EBDE: tdefl_compress_buffer (zip_file.hpp:2747)
==37967==    by 0x566506B: mz_zip_writer_add_mem_ex (zip_file.hpp:4430)
==37967==    by 0x5663B0F: mz_zip_writer_add_mem (zip_file.hpp:4190)
==37967==    by 0x566E63D: miniz_cpp::zip_file::writestr(std::string const&, std::string const&) (zip_file.hpp:5514)
==37967==    by 0x566E515: miniz_cpp::zip_file::write(std::string const&, std::string const&) (zip_file.hpp:5504)

I don't think I have the latest code of miniz-cpp so I also put the line here:

    }
    if (!dist) break; q = (const mz_uint16*)(d->m_dict + probe_pos); if (TDEFL_READ_UNALIGNED_WORD(q) != s01) continue; p = s; probe_len = 32;
    do { } while ( (TDEFL_READ_UNALIGNED_WORD(++p) == TDEFL_READ_UNALIGNED_WORD(++q)) && (TDEFL_READ_UNALIGNED_WORD(++p) == TDEFL_READ_UNALIGNED_WORD(++q)) &&
                   (TDEFL_READ_UNALIGNED_WORD(++p) == TDEFL_READ_UNALIGNED_WORD(++q)) && (TDEFL_READ_UNALIGNED_WORD(++p) == TDEFL_READ_UNALIGNED_WORD(++q)) && (--probe_len > 0) );
    if (!probe_len)
    {

Tracking the origin gives:

==44853== Conditional jump or move depends on uninitialised value(s)
==44853==    at 0x564B8A2: tdefl_find_match (zip_file.hpp:2338)
==44853==    by 0x564B8A2: tdefl_compress_normal (zip_file.hpp:2610)
==44853==    by 0x564C8F5: tdefl_compress (zip_file.hpp:2727)
==44853==    by 0x564CA9F: tdefl_compress_buffer (zip_file.hpp:2747)
==44853==    by 0x5652F2C: mz_zip_writer_add_mem_ex (zip_file.hpp:4430)
==44853==    by 0x56519D0: mz_zip_writer_add_mem (zip_file.hpp:4190)
==44853==    by 0x565C4FF: miniz_cpp::zip_file::writestr(std::string const&, std::string const&) (zip_file.hpp:5514)
==44853==    by 0x565C3D7: miniz_cpp::zip_file::write(std::string const&, std::string const&) (zip_file.hpp:5504)
...
==44853==  Uninitialised value was created by a heap allocation
==44853==    at 0x4A37ECB: malloc (vg_replace_malloc.c:307)
==44853==    by 0x56422AB: def_alloc_func (zip_file.hpp:1054)
==44853==    by 0x5652C34: mz_zip_writer_add_mem_ex (zip_file.hpp:4375)
==44853==    by 0x56519D0: mz_zip_writer_add_mem (zip_file.hpp:4190)
==44853==    by 0x565C4FF: miniz_cpp::zip_file::writestr(std::string const&, std::string const&) (zip_file.hpp:5514)
==44853==    by 0x565C3D7: miniz_cpp::zip_file::write(std::string const&, std::string const&) (zip_file.hpp:5504)
...
==44853== 

It would be great if you can solve this issue 😄

miniz-cpp is based on an old version of zlib/minizip and contains security vulnerabilities. See this example/failing test here: cleeus@d233931

I understand that this project is not in active development but it would be good to point this out in the README, otherwise this code might end up in critical code paths (and it probably alread has). There is a modern, well maintained version of minizip in https://github.com/nmoinvaz/minizip.

In line 518, mz_uint64_t is defined as unsigned long long.
However, std::uint64_t on my platform (Linux 64-bit Debian 11) is defined as unsigned long.

This leads to errors on multiple lines.

Here is the error :
zip_file.hpp:5611:30: error: invalid conversion from ‘std::size_t ()(void, uint64_t, const void*, std::size_t) {aka long unsigned int ()(void, long unsigned int, const void*, long unsigned int)}’ to ‘mz_file_write_func {aka long unsigned int ()(void, long long unsigned int, const void*, long unsigned int)}’ [-fpermissive]
archive_->m_pWrite = &detail::write_callback;

build and works well on macOS 10.14, clang v10

On line 5980 you use std::strlen. The namespaced version of this function is not defined in "string.h", only in <cstring>.

This leads to a compilation error.

So I want to add a save file to an archive. Because that file already exists it just "stacks" the same files on top of each other with the same filename. Is there a way to overwrite the existing file? I don't want to unpack the complete archive because that would take too long for just saving in the application.

Seems like all function declarations should be marked as static for correct linking from multiple translation units.

Could you provide an example on how to read actual data?

For example, use the zip command to compress a text file:

zip test.zip test.txt

And then your example would write on screen the test.txt original content.

Thanks

Would be nice to have a real documentation instead of guessing what method does what. :)

Is there a way to overwrite a file in a zip? Or delete it? Because when using writestr it just creates file with same name in the zip, which later obviously does not work correctly.

Calling "extractall" on a zip file doesn't create folders which exist in the archive, therefore not extracting file that exist within those folders. I would guess this is a limitation of fstream. For what it's worth, this was seen on Windows 7.

Hi,

Does this lib support MacOS symlink as zip command do (--symlink option) ?

Thanks