Hello, after last weekly update I found that some distfiles was packed with lzma (such as coreutils-6.10.tar.lzma, m4-1.4.11.tar.lzma and others). So I ask developers to add support of .lzma in deltup/getdelta.
There are very many lzma encoders out there which produce different lzma files for the same input file. Example: libpng-1.2.26.tar compressed with the lzma command line tool from the 7zip distribution using the command line lzma-$v e -si -so < tars/libpng-1.2.26.tar > libpng-1.2.26.tar.lzma-$v (where $v is the version) creates this files: -rw-r--r-- 1 root root 500691 23. Aug 14:32 libpng-1.2.26.tar.lzma-4.27 -rw-r--r-- 1 root root 500819 23. Aug 14:32 libpng-1.2.26.tar.lzma-4.43 -rw-r--r-- 1 root root 500819 23. Aug 14:32 libpng-1.2.26.tar.lzma-4.57 -rw-r--r-- 1 root root 500872 23. Aug 14:32 libpng-1.2.26.tar.lzma-4.63 -rw-r--r-- 1 root root 500872 23. Aug 14:32 libpng-1.2.26.tar.lzma-4.65 -rw-r--r-- 1 root root 500872 23. Aug 14:32 libpng-1.2.26.tar.lzma-9.20 So we have at least three different encoders with this one package. Trying lzma from lzma-utils-4.32.0_beta5 (with default settings) I even get another output file: -rw-r--r-- 1 root root 500404 23. Aug 23:18 libpng-1.2.26.tar.lzma-utils-4.32.0_beta5 So the problem of adding support for lzma is, that the server would have to test very encoders maybe with different settings and the clients would also need to have every encoder installed in order to be able to decompress all files. The bottom line is: It's very unlikely to get lzma support in deltup as long as we need the compressed checksums to match. I just filed a feature request on portage #380443 which addresses this problem. I'd say, as soon as that got implemented, the deltup server can just decompress the tar files in question, create the dtu file from them and let it up to the receiver to recompress the target file with it's compressor ignoring the compressed checksum alltogether.
Not attempting to obtain a diff at all for files ending with .xz or .lzma (and possibly their tarisms) would also be better than the current state.
Reassigning to maintainer-needed since package was not being maintained by anyone in tools-portage herd.
Apart from the problems mentioned in previous comments, this really is more of an upstream issue.