> Rich Teer wrote: > > On Wed, 23 Aug 2006, Alan Hargreaves wrote: > > > >> Oh if only it were that simple. Unfortunately it's > not. I've got code working > >> for producing the bits, I need to keep testing > each time a code drop happens > >> (and I really need to test sparc properly), but > other than that it's ready for > >> code review. > > > > Given that (presumably) the Solaris bits are > shipped in non-debug mode, > > can you please give us some more details as to why > building the open > > bits in non-debug mode causes so much grief? I > take it that it's not > > as simple as just changing an option in a > Makefile... > > It's not so much grief as it is time-consuming. > Here's the process my > cripts do for delivering the nightly builds each > week: > > 1. Bringover from /ws/onnv-gate > 2. Build a full (open+closed) build > 3. Move the proto area out of the way > 4. Setup the minimal required closed binaries > (libc_i18n & libdisasm on > sparc) > 5. Build an open-only build > 6. Build closed-binaries tarball (consists of the > following substeps) > 6a. Make a copy of the full proto area we saved in > n step 3 > 6b. Delete anything that is also in the open-only > y proto area > 6c. Delete anything we don't have the rights to > o redistribute > 6d. Pull in the RE-signed crypto bits > 7. Using the newly created closed-bins, fire off one > more open-only > build to build the BFU archives (so that the BFU > archives contain > the closed-bins) > reate the SUNWonbld package > 9. Create the BFU archives tarball > 10. Create the source tarball (removing all the SCCS > history) > 11. Generate changelogs > 12. Create the squelched source (consists of the > following substeps): > 12a. Bringover a child of usr/src > 12b. Squelch the history for every file > 12c. Upload to cvs.opensolaris.org > 13. Build the Mercurial bundles > > Steps 1-9 are done twice, once for SPARC, and once > for x86. > Essentially, they would have to be done two more > times - for non-debug > on both platforms. > > cheers, > steve > > > -- > stephen lau // [EMAIL PROTECTED] | 650.786.0845 | > http://whacked.net > opensolaris // solaris kernel development > _______________________________________________ > opensolaris-code mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/opensolar > is-code > as for the time-consuming part and given the number of servers that i can imagine would be involved with building solaris and OS, couldn't you partition just one server out to build the debug bits as is done currently, and dedicate another to do non-debug, then have them upload to the same mirror under (slightly) different names? it would be time consuming to initially set the second server up, but in the end would be a lot less time consuming than actually making one machine build both types of bits, since the closed bits would only be built once per architecture on each machine. this would also be easier on the admin, since he or she could login to both machines at once, issue commands to them, and not have to worry about sitting around for one build to complete. This message posted from opensolaris.org _______________________________________________ opensolaris-code mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
