I think this is better than setting ARCHIVE_MIRROR, which may be useful for other purposes--it's even more handy for managing our source archive if you can populate what's current with a global gmake download to one directory and then rsync that to dlc. As you say, with multiple workspaces, you can simply symlink this to a single copy of the archives, and with this done, you don't have to copy to pop into the workspace; you get archive dedup for free with nearly zero effort. The only downside is that userland-fetch might blow away your only copy because of digest mismatches in the component Makefiles, so you probably want to use snapshots a bit more than you might have done. People like me doing development on laptop SSDs get a big win for the reduced storage overhead.
On Sat, May 5, 2012 at 5:35 PM, Alasdair Lumsden <[email protected]> wrote: > Hi All, > > One improvement we added to s10-userland, was to download archives to a > single top level archives directory instead of into the component directory. > > We used: > > $(WS_TOP)/archives > > We did this by defining "USERLAND_ARCHIVES?= $(WS_TOP)/archives" in > shared-macros.mk and fixing up prep.mk. > > There are a few benefits to this, one is that if you have lots of > illumos-userland workspaces, they can all share a single archive directory. > It can be NFS or lofs mounted, and it makes rsyncing the contents of that > directory up to our mirror archive a lot easier than manually picking through > all the component directories. > > It also lets us add $(WS_TOP)/archives to .hgignore and no longer have to put > up with seeing downloaded archives in the "hg status" output. > > Before I begin work on this changeset for illumos-userland, does anyone have > any feedback? > > Cheers, > > Alasdair > _______________________________________________ > oi-dev mailing list > [email protected] > http://openindiana.org/mailman/listinfo/oi-dev _______________________________________________ oi-dev mailing list [email protected] http://openindiana.org/mailman/listinfo/oi-dev
