I second this On Fri, 25 Aug 2017 09:03:46 -0700 Chris Lamb <[email protected]> wrote: [...] > Hi, > > I was just building a large package from git-buildpackage with > pristine-tar and noticed that the creation of the source package > was taking rather a long time. >
Yes, in my case for gcc-arm-none-eabi it takes 4 minutes approx. > I looked at htop and noticed that only a single CPU (of the 8) was > being used in the call to `xz`, whilst xz's manpage itself claims > it can compress in "multithreaded" mode. > Me too, as far I can see you can pass by parameter how many threads do you want to use. > Would it be possible to enable this? :) I threw a naive extra > --threads=Sys::CpuAffinity::getNumCpus() and test_xz seemed to still > pass but I'm very likely missing something... > I've tried in a hackish way (my perl knowledge is almost null) adding the parameter `unshift @params, '-T4';` hardcoded in pristine-xz:genxz and I could see xz was using 4 cpus but the problem is that I got the following error: Pristine-tar couldn't verify "gcc-arm-none-eabi_6.3.1+svn253039.orig.tar.xz": pristine-tar: gcc-arm-none-eabi_6.3.1+svn253039.orig.tar.xz does not match stored hash (expected 8942af17cead0376934cd4934473b672f98caaefdf8334e5f82a65a853b4d485, got 24e6749a34b976672ed6398650dbede8fb15ecd7669a094170332b1fb64bfc06) So, using multi-threaded compression seems to result in a different binary :(. Also, is probable that I'm missing something, of course. -- TiN
signature.asc
Description: OpenPGP digital signature
