Date: Mon, 28 Jun 2004 21:26:42 +0400 From: Sergey Zaharchenko <doublef@tele-kom.ru> To: Robert Huff <roberthuff@rcn.com> Cc: questions@freebsd.org Subject: Re: Updating source code manually Message-ID: <20040628172641.GB2963@shark.localdomain> In-Reply-To: <16608.13888.194802.955936@jerusalem.litteratus.org> References: <BAY15-F11fWM5nWGsRw00004a38@hotmail.com> <20040628134639.GA5699@shark.localdomain> <16608.13888.194802.955936@jerusalem.litteratus.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--TakKZr9L6Hm6aLOc Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Mon, Jun 28, 2004 at 11:16:16AM -0400, Robert Huff probably wrote: >=20 > Sergey Zaharchenko writes: >=20 > > Just a minute. You shouldn't portupgrade KDE when KDE is running, > > but you should be able to run `make' to build everything while > > KDE is running, shut down KDE and `portupgrade -w' afterwards > > (which will use the binaries built by `make' and install them, > > taking seriously less time than the original `make') and restart > > KDE. At least that's how it would with an ordinary port. >=20 > One of the two of us is confused about this. > As I understand it: > A) Running "make build" but not "make install" doesn't really > solve the "installing while running" issue. Sure, it won't install > for that port, but it will build-and-install for every port upstream > ... I see. You are talking abiout ABI mismatch. I dug up a portion of your earlier post: > Let's say KDE uses libfoo.1.5.so, which is actually v1.5.7. > You upgrade something, which upgrades libfoo.1.5.so to v1.5.8. The > kernel (unaware of the change) reloads part of the file and restarts > execution at a particular address. Will that address valid code? When a file is open by a process, even if you unlink it and replace it with another one, the original file will stay on disk until the last file handle referencing it is closed. I assume that holds true for libraries too. So, EXISTING processes aren't screwed. But, when a NEW process is created, it references the NEW shared library, and if their ABI's don't match --- BOOM!:) That was just a correction. I was under the impression that the OP already had KDE (and thus all `father' ports) installed and up-to-date, and only wanted to patch a file. That would mean there would be 0 upstream ports rebuilt, or am I mistaken? Maybe my post looks like stating it's a universal approach. It isn't. Sorry I didn't mention it. But if the OP doesn't have KDE up-to-date, he could `downgrade' his ports tree to match his packages (reducing the problem to the previous one). Of course that's only possible if the patch applies to the `downgraded' KDE too. He will still have to build (not-up-to-date) KDE, but this should escape ABI worries, as in fact no other changes will be made. > B) ... unless you're suggesting starting at the top of the > dependency tree and doing build-but-don't-install by hand for every > component in order. I consider this severely impractical; it also ... The ABI problem isn't solved by your B) approach, which will only build `father' ports but still install `grandfather' ports. > C) ... won't work with portupgrade unless one uses the "w" option. Please read more carefully. I mentioned '-w'. --=20 DoubleF "When I was crossing the border into Canada, they asked if I had any firearms with me. I said, `Well, what do you need?'" -- Steven Wright --TakKZr9L6Hm6aLOc Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (FreeBSD) iD8DBQFA4FTRwo7hT/9lVdwRAm9NAJ4l3rV0IpKDWa5RYQn+fZEHMKET9QCfcmLU j0p0X4slXRuQcgdXjqwRw1E= =yaSa -----END PGP SIGNATURE----- --TakKZr9L6Hm6aLOc--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040628172641.GB2963>