On Mon, 28 Mar 2005 19:56:30 -0600 Scott Jones <[EMAIL PROTECTED]>
wrote:
| If you search gmail (it being that you save these mails) ...I am
| pretty sure you will find a post by Ciaran (damn dont rip my head off
| C. if I spelled your name wrong) that says it is bad and it all boils
| down to cache.  So I wouldn't do it but if you really want to
| understand why ask ciaran and risk his wrath...

Noooo, cache is bad for sync + merge at the same time. Parallel merge is
bad for different reasons.

Simple example, which is easy to understand but not entirely valid since
we generally check for this kind of thing...

Existing state: libfoo-1 is installed.
User:           merge libfoo
User:           merge fnord
libfoo:         dep resolver decides to upgrade to libfoo-2
fnord:          dep resolver decides to install fnord
libfoo:         fetch, unpack
fnord:          fetch, unpack
libfoo:         configure starts
fnord:          configure starts
libfoo:         configure done, make starts
fnord:          configure: check for libfoo: found libfoo-1
fnord:          configure done, make starts
fnord:          make building with libfoo-1 headers
libfoo:         make done, installing
libfoo:         install done
libfoo:         autocleaning libfoo-1
libfoo:         merge complete
fnord:          linking against libfoo-1

And then stuff explodes, since libfoo-1 isn't there any more. The dep
resolver won't necessarily see this as a problem either, assuming libfoo
isn't slotted and that fnord doesn't need specific libfoo versions.

Usually the breakages are quite a bit more complex than that.

-- 
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail            : ciaranm at gentoo.org
Web             : http://dev.gentoo.org/~ciaranm

Attachment: pgpCi8AaKMT3m.pgp
Description: PGP signature

Reply via email to