[05:39:00] --- jaltman has left: Replaced by new connection [05:39:01] --- jaltman has become available [06:44:04] --- jaltman has left: Disconnected [06:44:16] --- jaltman has become available [07:03:32] What is the recommended way to use git on Windows? [07:03:57] Well, first you gouge out both of your eyes. [07:05:04] --- matt has become available [07:05:07] If I'm running Windows in a virtual machine, would it be preferable to mount a repository that I've checked out on the host? [07:05:22] The issue is line ending fun. [07:05:33] Right... [07:05:46] I would speak directly to Jeffrey Altman and find out what he's doing. [07:05:53] (Or Asanka) [07:06:47] Last time I helped Jeffrey with any git stuff, I think he was using msysgit. But I think his repository was also corrupting itself on a regular basis. [07:07:00] Bypass could work on 2.4, but at some point we assumed you'd use it only if you had readpages (2.5+). Perhaps that's arbitrary. [07:08:08] He says you must only use mysgit. I only use cygwin. I have learned to deal with the posix acls "integration" "help" it now provides. But no corruption ever. [07:08:15] Will the MS compiler read LF line breaks or would I have to transcode everything to CRLF if I'm only reading the code for compiling? [07:08:30] matt: But you'd have no issue if we just remove the current bypass code from the LINUX24 tree? [07:08:51] (I think we've already done so, so that's a bit of a retroactive question. But hey, we could always put it back) [07:09:06] it's dead, jim [07:09:23] No, I don't think so. It occurred to me that we might use it on a platform that was more like 2.4--but we can deal with that later. It's a good idea for now. [07:09:49] the code that was in the 2.4 tree appeared to be (mostly) the 2.6 code anyway [07:10:02] Correct. [07:10:25] In other news, yet more breakage from the RPMFusion RPMs - they've changed where all of the include files live. [07:23:57] --- deason has become available [07:49:16] --- reuteras has left [08:10:03] --- abo has left [08:26:54] JSund: For OpenAFS at least you want to use a repo that is located within the VM because you need to configure git to always convert LF to CR-LF. Not all of the Windows tools will do the right thing with just a LF. [08:27:56] The git repository corruption was caused by a handful of files in the git repo that were CR-LF when they should not have been. [08:37:15] jaltman: I think I got that part correctly with msysgit after a while. [08:37:55] If either of you fancy writing something about how to use git on Windows to work with the OpenAFS tree, it would be great to add that to the existing documentation. [08:38:22] we'll have a new wiki soon, so you may want to hold off on adding it to the old one [08:39:03] I've only look at my own git repository at the moment and I'll probably do as little as possible of the git work in Windows at the moment. [08:39:14] looked at* [09:11:42] JSund: A trick I've used for many CMS tools (but not git so far) is to run in a VMWARE machine in the appropriate OS and use HGFS to bring the repo onto windoze. You get the tool where it is most stable, but you still need to be *very* careful about cr/lf and tabs [09:15:33] rod: Would that be simliar to mounting a directory as a network mount in VirtualBox? [09:15:44] yea [09:15:55] the challenge with that approach is that if you ever end up mixing tools that create LF and CR-LF on the same system you end up in trouble. [09:17:41] the problem I was running into was that two of the Unix only source files in the repo had some stray lines with CR-LF because they must have been edited on a Windows system years before the code was released as OpenAFS. It was never an issue with cvs but with git the sha1 hashes would not compute correctly and the repo was corrupted. [09:19:50] [09:27:57] --- Russ has become available [09:30:35] --- jaltman has left: Disconnected [09:36:05] --- jaltman has become available [09:56:46] --- Kevin Sumner has left [09:57:46] --- rod has left [10:22:29] --- jaltman has left: Disconnected [10:28:32] --- Kevin Sumner has become available [10:35:02] --- jaltman has become available [11:10:27] --- tkeiser has become available [12:08:14] --- pod has left [13:03:22] --- jaltman has left: Replaced by new connection [13:03:23] --- jaltman has become available [13:16:18] --- jaltman has left: Disconnected [13:16:20] --- jaltman has become available [13:33:06] --- jaltman has left: Disconnected [13:53:33] --- tkeiser has left [13:54:09] --- tkeiser has become available [14:09:09] --- pod has become available [14:15:06] --- mdionne has become available [14:25:17] --- tkeiser has left [14:26:51] --- pod has left [14:30:38] --- pod has become available [14:32:18] --- kaj has become available [14:58:40] --- tkeiser has become available [15:05:52] --- kaj has left [15:05:52] --- kaj has become available [15:06:49] --- tkeiser has left [15:07:14] --- jaltman has become available [15:08:14] --- jaltman has left: Replaced by new connection [15:08:15] --- jaltman has become available [15:17:27] --- kaj has left [15:19:42] --- kaj has become available [15:49:20] --- deason has left [16:30:29] --- JSund has left [17:55:20] the wiki has been replaced and there's a new link to it on the web site, fwiw [18:43:32] --- matt has left [19:08:00] --- mdionne has left [19:09:18] Hm. FBSD kernel compiles with -mcmodel=kernel. gcc(1) says: -mcmodel=small Generate code for the small code model: the program and its symbols must be linked in the lower 2 GB of the address space. Pointers are 64 bits. Programs can be statically or dynamically linked. This is the default code model. I ... kind of want to rewrite MakefileProto.FBSD.in to .include [19:09:31] Ew, sorry about the wrapping. [19:29:16] > MakefileProto.FBSD.in to .include [19:29:27] that's probably reasonable. we work with linux kbuild for exampke [19:32:19] I'm somewhat unconvinced that this explains my memory corruption issues, though. [19:35:18] it might. red zone or no? [19:35:51] --- kaj has left [19:36:17] kernel is compiled with -mno-red-zone, but we aren't. [19:37:07] "oops" [19:37:21] and you're on amd64. uh. might wanna fix that [19:38:03] maaaaybe ... [19:38:44] Do I just want that in MakefileProto, you think? [19:39:49] probably. [19:39:52] well, .in [19:40:10] Well, yes. [20:09:07] Damn. That appears to have been it. [20:28:09] --- kaj has become available [20:28:17] yeah. see. had that fun with linux when i worked on the amd64 port there [20:28:29] I see. [20:29:09] Hm, now I have a deadlock. "This should be easy to debug, right?" [20:29:24] no worse than before [20:35:14] --- jaltman has left: Replaced by new connection [20:35:15] --- jaltman has become available [20:49:43] --- kaj has left [21:01:14] --- deason has become available [21:28:25] --- Born Fool has become available [21:35:37] We have conditionals on arch in MakefileProto.FBSD.in; where is the set of valid things to condition on defined? [21:36:34] (And if they're the sysnames as I suspect, I bet I can only list them one at a time ...) [22:13:39] --- reuteras has become available [22:39:19] --- Born Fool has left [22:46:10] --- deason has left [23:32:20] --- Russ has left: Disconnected