Shell

I have some testing for deflate (zip/gz) compression speed was increased by 1-3% I have compressed a big (298 GiB, including ~48 GiB sparse ranges) file to GZip format with 7zG 21.07 and 25.00. I have used -mx9 -mmt- in all scenarios, while varying fb and pass. -mfb256 -mpass1 have negligible difference 05:19 vs 05:20 total time, rounded to the nearest minute). -mfb258 -mpass1, on the other hand, performs better on about 1% with 25.00 (09:42 vs 09:48). Of course, justifying 1% compression speed improvement...

If you need to ignore any files in the existing archive, you should use -u!p0q0r2x2y2z2w2. Usually, you can shorten this switch by taking into account the default values. For instance, 7z a -u!p0q0 is enough. If the new archive name corresponds to an existing file that is not a vaild archive of the same type, you should delete it manually. 7-Zip currently does not have an option not to analyze the previous archive. It is strange, though, that 7-Zip leaves incomplete archives. Usually, this happens...

You can create your archive in two passes. For the first time, you exclude files you don't want to compress. For the second time, you include those files and specify compression method Store. In command line it might look as follows: 7z a -mx9 -x!*.mp4 files.7z * 7z a -mx0 -i!*.mp4 files.7z There are also third-party solutions to do so. For example, there is a plugin for 7-Zip that applies different compression to pre-defined groups of files.

Like some other archivers, 7-Zip utilizes some filters that improve compression of certain files. For example, CAB and RAR archives use filters for x86 code similar to BCJ/BCJ2. RAR archives can use Delta filter that is probably identical to that in 7-Zip. Unlike other archivers, 7-Zip allows specifying almost any combination of filters and compression methods. For example, it is possible (but not very useful) to apply Delta filter, then compress with Deflate, and then compress again with LZMA2....

7za.exe 23.00 x64 (CRC32: 61509E65) does include two calls to RtlGetVersion(), both times with calling GetProcAddress(). The first case, IMAGE_BASE+0xC793, must belong to IsItWindowsNT(), though this address is not actually executed for the "List archive" command. Version 22.01 calls GetVersionEx() instead - it is mentioned in CPP\Windows\NtCheck.h. Maybe you have forgotten to delete some more checks so this code gets compiled. I do not deny there must be reasons for changing the code - in fact,...

I am examining 7-Zip standalone x64. The code for IsItWindowsNT() is somewhat different in the new version. Why do you use RtlGetVersion() instead of Win32 API's GetVersionEx()? And you seem to omit initialization of dwOSVersionInfoSize. However, Microsoft guys seem to write basic _OSVERSIONINFO regardless of the size member, so your code works as intended.

But the the rules for pair archive_file-disk_file are only for full paths: archive:: folder1\file1 vs disk:: folder1\file1 OK, I understand now. According to your comment, I can do recursive deletion of multiple files only with a listfile and without -uq1, so that the matches are taken not from the disk.

Attached a sample archive (both WIM and 7z). Suppose file1 is present in current directory. I want to remove all files with the same name from the archive (there are 3 of them). 1. 7z d -r -u- -uq1!test2.7z test.7z * removes only one file1 - the one in the root directory of the archive. 2. 7z d -r -u- -uq1!test2.7z test.7z file1 does exactly the same. 3. 7z d -r -u- -uq1!test2.7z test.7z -ir!file1 does the same. 4. 7z d -r -u- -uq1!test2.7z test.7z *\file1 deletes nothing. For WIM, these commands...