Le Tue, Jan 17, 2023 at 09:28:18AM +0000, Stuart Henderson a écrit :
> On 2023/01/16 15:30, Christian Weisgerber wrote:
> > CVSROOT:    /cvs
> > Module name:        ports
> > Changes by: [email protected]   2023/01/16 15:30:55
> > 
> > Modified files:
> >     archivers/xz   : Makefile distinfo 
> >     archivers/xz/patches: patch-config_h_in 
> >     archivers/xz/pkg: DESCR 
> > Removed files:
> >     archivers/xz/patches: patch-src_xz_main_c 
> > 
> > Log message:
> > archivers/xz: update to 5.4.1
> > 
> > Significant changes:
> > * threaded .xz decompressor
> > * decompression support for the .lz (lzip) file format
> > * upstream pledge(2) sandboxing
> 
> Would it make sense to switch lzip decompression to use this instead?
> xz is much more widely used in ports, so this avoids having to install
> an extra package as BDEP when you occasionally work on a port with lz
> distfiles. (Also fixes a small bug where lz patch files don't work
> unless the main distfile is .tar.lz).
> 
> I'm also wondering if we should add a way to pass user-defined flags to
> xz; the new threaded decompressor saves a lot of clock time extracting
> large xz files (including ports like chromium, firefox, texmf).
> (use xz --decompress instead of xzdec, and add a new variable that the
> user can set in mk.conf added to the command; -T0 to use all available
> cores or use some specific other value).

id welcome such a thing:

with xz 5.2:
===>  Extracting for firefox-109.0
   10m15.13s real     1m18.73s user     3m09.43s system

with  xz 5.4, using xz --decompress -T0
===>  Extracting for firefox-109.0
    6m48.21s real     1m20.75s user     3m32.08s system

Reply via email to