[gentoo-dev] Remove dev-status of mips profiles
Can we please move the mips profiles from dev to exp in profiles/profiles.desc? The ~150 mips development profiles increase the time for a `repoman -d full` run in dev-perl/ from three to five minutes. That is an increase of roughly 66 percent. repoman further prints more than 2000 lines of output for two keywording problems. I can't remember any serious mips keywording activity in the last years. I don't see why I should spend any more time on checking their profiles and filing bugs for an unactive team. Open KEYWORDREQ bugs: https://bugs.gentoo.org/buglist.cgi?quicksearch=kw%3AKEYWORDREQ+owner%2Ccc%3Amips%40gentoo.org
Re: [gentoo-dev] Remove dev-status of mips profiles
On Thu, Feb 11, 2010 at 04:54:38PM +0100, Torsten Veller wrote: Can we please move the mips profiles from dev to exp in profiles/profiles.desc? The ~150 mips development profiles increase the time for a `repoman -d full` run in dev-perl/ from three to five minutes. That is an increase of roughly 66 percent. repoman further prints more than 2000 lines of output for two keywording problems. Quick pcheck visibility scan of the full tree, stats follow: mips profiles still enabled: * 116191 seperate dependency issues, 1 line per profile/dependency issue * roughly 2m39s run time mips profiles disabled (leaving mips-irix however) * 9550 seperate dependency issues, 1 line per profile/dependency issue * roughly 1m54s run time. So... mips accounts for about 30% of the pcheck runtime, and *92%* of known visibility issues. As for the runtime difference between pcheck/repoman, pcheck has some tricks internally to reduce the # of profiles it has to scan down to just the unique USE/mask set- I'd expect the mips impact to be far larger w/out that trick in place. At the very least if it's going to be kept around, experimental or not, the number of profiles in use there *really* needs reduction- mips has roughly 117 profiles listed in profiles.desc out of 217- literally ~54% of all dev/stable/experimental profiles. Either way, stats to chew on. ~harring pgpSRzP1oftUl.pgp Description: PGP signature
Re: [gentoo-dev] Remove dev-status of mips profiles
On Thu, 11 Feb 2010 10:41:41 -0800, Brian Harring ferri...@gmail.com wrote: On Thu, Feb 11, 2010 at 04:54:38PM +0100, Torsten Veller wrote: Can we please move the mips profiles from dev to exp in profiles/profiles.desc? The ~150 mips development profiles increase the time for a `repoman -d full` run in dev-perl/ from three to five minutes. That is an increase of roughly 66 percent. repoman further prints more than 2000 lines of output for two keywording problems. Quick pcheck visibility scan of the full tree, stats follow: mips profiles still enabled: * 116191 seperate dependency issues, 1 line per profile/dependency issue * roughly 2m39s run time mips profiles disabled (leaving mips-irix however) * 9550 seperate dependency issues, 1 line per profile/dependency issue * roughly 1m54s run time. So... mips accounts for about 30% of the pcheck runtime, and *92%* of known visibility issues. As for the runtime difference between pcheck/repoman, pcheck has some tricks internally to reduce the # of profiles it has to scan down to just the unique USE/mask set- I'd expect the mips impact to be far larger w/out that trick in place. At the very least if it's going to be kept around, experimental or not, the number of profiles in use there *really* needs reduction- mips has roughly 117 profiles listed in profiles.desc out of 217- literally ~54% of all dev/stable/experimental profiles. I agree, I wasn't sure why so many profiles were added[1] for a dead team (for all intensive purposes). Seems quite silly to me to leave them as 'dev' status. If a member of the mips team would reply to this thread, that would be good. (and surprising to me :) I would guess that it would be far easier to work in an overlay at this point. I would also guess that if there are ANY mips users out there that they would have to use some other ACCEPT_KEYWORDS value because the shape of ~mips is so...bad. -Jeremy [1]: http://sources.gentoo.org/viewcvs.py/gentoo-x86/profiles/profiles.desc?r1=1.151r2=1.152 Either way, stats to chew on. ~harring
Re: [gentoo-dev] Remove dev-status of mips profiles
On Thu, Feb 11, 2010 at 06:56:58PM +, Jeremy Olexa wrote: I would guess that it would be far easier to work in an overlay at this point. I would also guess that if there are ANY mips users out there that they would have to use some other ACCEPT_KEYWORDS value because the shape of ~mips is so...bad. Might want to wait till the cvs-git transition occurs, assuming it's in the next few months- primarily, preserve history, they can just fork the tree off and mainline can start pruning. Till then the profiles could pretty easily be disabled if folks wanted. ~harring pgp35n9LaxOK7.pgp Description: PGP signature
Re: [gentoo-dev] Remove dev-status of mips profiles
On Thu, Feb 11, 2010 at 06:56:58PM +, Jeremy Olexa wrote: At the very least if it's going to be kept around, experimental or not, the number of profiles in use there *really* needs reduction- mips has roughly 117 profiles listed in profiles.desc out of 217- literally ~54% of all dev/stable/experimental profiles. Correction also- they've got 154 rather than 117, meaning 71% of all profiles for a scan. Not sure how I screwed that number up, but aparently I managed it. ~harring pgpJ6dBz0WF9a.pgp Description: PGP signature
Re: [gentoo-dev] Remove dev-status of mips profiles
On 02/11/2010 08:56 PM, Jeremy Olexa wrote: Can we please move the mips profiles from dev to exp in profiles/profiles.desc? The ~150 mips development profiles increase the time for a `repoman -d full` run in dev-perl/ from three to five minutes. That is an increase of roughly 66 percent. repoman further prints more than 2000 lines of output for two keywording problems. Quick pcheck visibility scan of the full tree, stats follow: mips profiles still enabled: * 116191 seperate dependency issues, 1 line per profile/dependency issue * roughly 2m39s run time mips profiles disabled (leaving mips-irix however) * 9550 seperate dependency issues, 1 line per profile/dependency issue * roughly 1m54s run time. [ .. ] [1]: http://sources.gentoo.org/viewcvs.py/gentoo-x86/profiles/profiles.desc?r1=1.151r2=1.152 We could at least delete the server/desktop/developer profiles from the above list right away, not really required to have them for every machine type IMHO