Summary: | bootstrap.sh hardcodes CONFIG_PROTECT="-*" | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Oleh Kravchenko <oleg> |
Component: | [OLD] Core system | Assignee: | Gentoo's Team for Core System packages <base-system> |
Status: | VERIFIED INVALID | ||
Severity: | normal | CC: | gentoo-bugs, zmedico |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | emerge --info |
Description
Oleh Kravchenko
2009-02-13 09:40:12 UTC
This is likely unrelated to etc-update, but to Portage's internal config handling. Check the output of "emerge --info | grep PROTECT" CONFIG_PROTECT tells portage to not automatically replace files in the given directories, so they get handled by etc-update or dispatch-conf. CONFIG_PROTECT_MASK allows you to then tell Portage that certain locations within CONFIG_PROTECT should have their files automatically replaced (so they don't get handled by etc-update or dispatch-conf). If you're still having problems, please post the full output of "emerge --info" and describe which options you're using in etc-update. The real problem is Portage's --noconfmem (or the absence thereof in default EMERGE_DEFAULT_OPTS) as it causes necessary (e.g. incompatible) config file changes to be disregarded. I did never care to look, if this feature simply buggy or conceptually broken, but changed it in make.conf locally. A proper version of a feature to deal with config files changes transparently would store a hash of the installed "vanilla" config files sans comments and do a comparison, if the config file (again: sans comments) coming with a newer ebuild differs to it (instead the user modified config file) to deduct, if it may be necessary to pester a human with it. (In reply to comment #0) > When I updating base packages, configuration in /etc/conf.d set to default :( As said in comment #1, please post CONFIG_PROTECT and CONFIG_PROTECT_MASK settings. Perhaps you've got CONFIG_PROTECT="-*"? (In reply to comment #2) > The real problem is Portage's --noconfmem I haven't seen anything in comment #0 to suggest that --noconfmem is relevant to this bug, since confmem deals with things not being merged rather than config protection. Created attachment 181998 [details]
emerge --info
Overwriting /etc/conf.d files happens when I call bootstraping $ /usr/portage/scripts/bootstrap.sh $ emerge -e world (In reply to comment #5) > Overwriting /etc/conf.d files happens when I call bootstraping > > $ /usr/portage/scripts/bootstrap.sh > $ emerge -e world > I don't know where you got the idea to call bootstrap.sh, but it has CONFIG_PROTECT="-*" hardcoded internally. Since bug #247776, there's already a warning produced by emerge when CONFIG_PROTECT is empty. Reassigning to base-system in case they want to change the behavior of bootstrap.sh with respect to CONFIG_PROTECT. i dont see the point. dont run bootstrap.sh on a stage3+. if it breaks, you get the pieces. I am porting gentoo on my pda. So i need bootstrap :) details about pda here http://www.kaa.org.ua/ru/asus-p535/hardware.html that is irrelevant. bootstrap is run natively on the device by catalyst. if you run it by hand, you better know what you're doing (including clobbering of config files). |