[03:19:09] I'm just looking at the perl-AFS stuff now. [04:07:17] --- Stephan Wiesand has become available [04:29:50] perl-AFS 2.63 is sufficiently different from 2.6.2 that the perl-AFS stuff isn't going to be an easy backport. [04:37:23] But I have something that "works". I'll push it and we can discuss whether it just bloats libafsauthent_pic.a too much. [05:00:24] --- Derrick Brashear has left [05:39:04] 8808-8811 - and perl-AFS 2.6.3 builds and runs with a few minor patches [05:41:56] the required make changes seem to cause trouble [05:42:10] The ones Andrew pushed last night? [05:42:40] yes, 8799 and 8800 [05:43:39] Yeah. One of them triggers a bug with a missing dependency. The other isn't pulled up correctly. [05:43:52] (AFS_CCRULE takes an argument, the older CCRULE doesn't) [05:44:31] did you fix it in your last push? [05:44:49] Yeah, both of those should be fixed. We'll see what buildbot says ... [05:45:35] is buildbot building with -j on vms with a single virtual core? [05:45:54] I don't know the details of the buildbot VMs, unfortunately. [05:46:04] I build with -j8 here, on a machine with 8 cores. [05:48:08] I'm just preparing a build including anything possibly targeted for pre2. This will run on a couple of different systems with -j2/4/8/16. Since you say it should work, I'll include the whole stack. [05:49:55] Did 1.6.1 or 1.6.2pre1 work in that environment? I think a lot of the parallel build work that Marc did happened on master, and I'm not sure how much of it got pulled up [05:51:05] Yes, all 1.6 has built fine in parallel for me. [05:51:42] That's why I would consider it a regression if it no longer worked with pre2. [05:52:43] You had objections to changing the 2.4 code in the "Avoid infinite d_invalidate loop" change - is this still true with the "simple fix" we agreed on? [05:54:53] I'm happy with the simple fix going to both places. [06:15:27] --- Marc Dionne has become available [06:22:36] Stephan: 8810 is missing a ; in the symbol list. I'm about to push a new change (I didn't see it on Mac OS, because we trim the semicolons there) [06:39:13] Thanks. Grabbed the new 8810 and 8811. Userland build -j16 just succeeded. [06:43:58] Is there really no way to parallelize configure? [06:46:32] You'd think that there should be, but I think the real problem is that it's just a shell script. [06:48:49] Amdahl's law is cruel when you build the kernel module on a vanilla farm node these days. [06:51:51] The problem is all of the kernel feature tests. [06:52:11] If vendors didn't cherry-pick changes, we could just use kernel version numbers. But they do, so we can't. [06:54:08] No way, I agree. [06:58:11] Here's the "bloat report": [07:00:00] Before: 200K./libafsauthent.so.1.1 676K./libafsrpc.a 68K./libafscp.a 692K./libafsrpc_pic.a 12K./libkopenafs.a 320K./libafsrpc.so.1.5 436K./libafsauthent_pic.a 420K./libafsauthent.a 8.0K./libkopenafs.so.1.1 [07:00:27] after: 1.3M./libafsauthent.so.1.1 676K./libafsrpc.a 692K./libafsrpc_pic.a 12K./libkopenafs.a 1.2M./libafsrpc.so.1.5 880K./libafsauthent_pic.a 748K./libafsauthent.a 20K./libkopenafs.so.1.1 [07:03:22] Shared object from 200K to 1.3M. Wow. [07:03:41] Not that anything we ship actually uses the shared object... [07:04:30] Actually, something wierd is going on with your numbers - libafsrpc.so goes from 320K to 1.2M, and there's no new code in there. [07:06:39] 8809 is in there too [07:07:09] That just adds two exports - the code is already present. [07:07:18] And libkopenafs.so goes from 8K to 20K [07:08:43] Wait, the first build may have used different options. [07:11:38] sorry. these are the correct "before" figures: 728K./libafsauthent.so.1.1 676K./libafsrpc.a 692K./libafsrpc_pic.a 12K./libkopenafs.a 1.2M./libafsrpc.so.1.5 436K./libafsauthent_pic.a 420K./libafsauthent.a 20K./libkopenafs.so.1.1 [07:12:59] Okay, so the libafsauthent.so and _pic libraries approximately double in size. [07:13:30] No big deal IMHO. [07:14:02] Yeah, I think it's fine - especially if it means we can finally get to the point where perl-AFS works properly :) [07:30:13] The client works, too :) [07:32:17] If you give me the patched perl-AFS source, I may get someone here to test it with this build. [07:36:36] It's a mess - I haven't made anything conditional on AFS version, but happy to email it through to you. [08:08:55] --- deason has become available [08:31:53] --- Stephan Wiesand has left [09:34:14] --- stephan.wiesand has become available [11:06:55] --- stephan.wiesand has left [14:25:14] --- kaduk@mit.edu/barnowl has become available [15:03:42] --- deason has left [15:12:54] --- kaduk@mit.edu/barnowl has left [15:13:03] --- kaduk@mit.edu/barnowl has become available [15:13:10] --- kaduk@mit.edu/barnowl has left [15:13:39] --- kaduk@mit.edu/barnowl has become available [19:01:32] --- Marc Dionne has left [21:47:29] --- kaduk@mit.edu/barnowl has left: Lost connection [21:49:53] --- kaduk@mit.edu/barnowl has become available [22:08:59] --- Derrick Brashear has become available