[04:03:06] [connected at Thu Mar 2 04:03:06 2006] [04:03:06] <> *** Looking up your hostname... [04:03:06] <> *** Checking ident [04:03:11] <> *** Found your hostname [04:03:37] <> *** No identd (auth) response [04:03:37] <> *** Your host is clarke.freenode.net[clarke.freenode.net/6667], running version hyperion-1.0.2 [04:03:37] [I have joined #peak] [09:00:33] ** erikrose has left IRC () [09:10:30] ** whit has joined us [09:19:59] ** erikrose has joined us [12:39:54] ** pje has joined us [13:44:52] pje is now known as pje_mtg [14:17:42] pje_mtg is now known as pje [16:35:53] ** DesreveR has joined us [16:51:24] ** r0n|mess has left IRC (Connection timed out) [17:09:59] ** bitmonk has joined us [17:10:21] pje: what's up with pyprotocols and zope interfaces? esp z3/z2 cross-compat.. [17:10:55] what's up is I haven't done anything with that in a few years. :) [17:11:06] ;) [17:12:09] so i imagine there has been limited interest in using this func? [17:12:19] pje: right now peak/pyprotocol breaks zope.interface where it tries to patch adapt [17:12:22] iirc [17:12:35] 0.9.3, or the trunk? [17:13:28] pje: whit said latest .egg a min ago [17:13:59] patches are welcome [17:13:59] pje...let me check [17:14:12] :) [17:14:20] pje: I'm not sure how to fix it, just how to disable it :) [17:14:38] don't figure that is the kind of patch you want.... [17:15:09] * whit will take a look when he has a chance though [17:15:22] Well, at least I'd know what's breaking. :) [17:16:12] * bitmonk might poke around [17:16:31] one sec....let me dig [17:16:44] luckily I caused this error this morning [17:21:37] almost there [17:23:23] DesreveR is now known as r0nny [17:25:02] pje: it's in protocols.zope_support (peak .5a3) [17:25:27] basically, I just stop pyprotocol from patching zope.interface [17:25:34] yo [17:25:43] I'm probing it right now to get more info [17:25:50] 0.5a*3*? [17:26:07] yes [17:26:42] Zope 2.9 / zope 3.1 [17:26:47] That's over 2 years old. [17:27:01] Use the separately-distributed PyProtocols package. [17:27:16] thanks [17:27:32] Let me know if the problem still exists, or is broken in some other way. [17:27:39] is there a better version of peak to use? svn? [17:27:50] * whit just did this with the easyinstall [17:28:10] easy_install -f http://peak.telecommunity.com/snapshots/ PEAK [17:28:23] Or check out from SVN, yeah. [17:28:26] pje: sorry to bother, I have a user trying to make a gentoo package using setuptools, and he is having some problems which I can't work out [17:28:29] I should probably add a dev link [17:28:29] cool! [17:28:33] thanks [17:28:34] r0nny: can you explain please? [17:31:44] hey pje, "easy_install -f http://peak.telecommunity.com/snapshots/ PEAK" grabs 5a3 [17:31:56] * whit goes to try from svn [17:32:28] well - if i do a egg based install everything works fine, but if i do a file based install it throws all data like icons/pixmaps to /usr/{icons,pixmaps,...} [17:33:01] * pje thinks he answered this question already on distutils-sig [17:33:17] just in the last couple days, actually. [17:33:29] it was my mail [17:33:42] right, well, what if it is non-package data? [17:33:47] bit i fail to understand, what exactly u mean [17:34:05] I mean, I could put it in a package if it has to be there [17:34:33] if it's not package data, it has no guaranteed install location. [17:34:49] oh, then I'll make it packae data [17:34:50] if you care where it goes, don't use data_files [17:35:14] that goes for the distutils too, setuptools doesn't change the distutils' handling of data_files [17:35:16] pje: I don't care where it goes, but I do care that I can access it using the resource api at runtime [17:35:34] If it's non-package data, it can go in the .egg-info directory, in which case the metadata API can find it. [17:35:58] However, it's rare that what you have is really project-level data, as opposed to data that belongs to some package in the project. [17:36:21] like icons and glade files and things? [17:37:23] can I get external data and mark it as package data using setup.py? [17:37:58] whit, I don't understand why your easy_install finds 0.5a3; mine finds the right version. Do you have an --allow-hosts mask? [17:38:22] I don't thinks so [17:38:25] aa_, your two choices for data are to put it under your .egg-info, or in a package directory. [17:38:39] (if you want to be able to find it at runtime) [17:38:52] ok, thanks [17:39:13] and the package directory is the recommended option in most cases [17:40:23] ok, well, it will grab a directory recursively for me, I barely need to do anything [17:40:48] ** plmxn has joined us [17:41:07] ** plmxn has left us [17:42:07] pje: platform...I'm on a mac [17:42:55] whit, there are source distros in /snapshots, it should still work. I tried from cygwin, and it d/led source [17:43:19] hmmm [17:43:54] ah [17:45:34] pje: could u add a shared data feature for stuff like icons/glade-files so they can be placed in /usr/share/ on package-manager based installs ? [17:45:38] pje: thanks....that works great [17:46:04] r0nny, if you just want it to be shared data, use data_files [17:46:25] You simply can't guarantee where those files will be at runtime, because users can customize the install_data location [17:46:46] pje: when we used data-files it was thrown over /usr, and had no runtime-determinable location [17:46:48] You just can't have it both ways: either it's a customizable shared location, or it's not [17:47:37] By definition, packaging systems control what prefix stuff goes under. [17:47:58] For example, on many BSD systems, stuff goes in /usr/pkg subdirectories, e.g. /usr/pkg/share [17:48:25] Ergo, there is no safe way to determine the runtime location, especially since users can set install_data as an individual option. [17:49:04] You can't have your cake and eat it too: either the data goes in a system-defined location or you define where it goes. [17:51:57] how about having them go to $pythonlib/site-packages/pkgdata/ [17:58:51] That would require all the packagers to agree on that convention. [17:59:07] As it is, some of them are not fond of setuptools installing any data under site-packages at all. :) [17:59:40] I'm still confused as to why you can't put the data in a package or in .egg-info, anyway. [17:59:55] If you need runtime access to it, that's where it should go. [18:00:29] i think ist more clean from the pkg-management point of view to have the data at a global shared position [18:00:53] Quite the opposite, in fact. [18:01:16] That makes it impossible for users to install private packages. [18:01:43] Global locations are anathema to user-installation of software. [18:02:37] usually a user has no access to the pgk manager, and casue of this hes using egg files [18:02:49] right. [18:03:00] So the data should be in the egg, not a global location. [18:03:10] i.e., in the .egg-info directory, or in a package directory. [18:03:25] So I'm still not understanding your question or problem. As far as I can tell, there is no problem. [18:03:35] no, I am switching the code over [18:04:07] I can't work out what path to find the new stuff under though [18:04:38] aa_, if it's package data, use pkg_resources.resource_string('packagename','subdir/file') [18:04:50] or resource_stream, or resource_filename if you really really need a filename. [18:04:56] ok, thanks [18:05:09] paths to resources should be /-separated; it will be replaced w/platform separator automatically [18:05:09] I was using Requirement.parse() for everything [18:05:24] ah [18:09:19] the problem with distutils data files was it was possibel to locate it while using egg, but it broke as soon, as it was installed pkg-manager friendly - maybe it would be usefull, to move all this data to the .egg-info dir and make it aviable via Requirement.parse() [18:18:31] pje: btw - i have another problem - the egg build fails, if i dont have the file /usr/lib/python2.4/site-packages/site.py [18:18:34] pje: looks like all works now...thanks for your help! [18:19:48] r0nny, what do you mean by "fails"? [18:20:12] whit: so pyprotocols and zope.interface in z2/3 are kosher? [18:20:35] bitmonk: if you install from the snapshots, all is well [18:20:47] yay, it is working so far. thanks pje [18:20:48] eenterestink.. [18:20:54] just need to change a lot of stuff [18:21:30] pje: then it stops the install, and deletes the generated files [18:21:50] okay, what do you mean by "the egg build"? [18:22:24] Are you talking about "setup.py bdist_egg"? [18:22:26] right after creating the egg file it stops with an error, and wont install the egg file to the system [18:22:28] Or something else? [18:22:55] Are you using easy_install or running a setup.py? [18:23:10] running setup.py [18:23:28] okay, so are you doing bdist_egg, install, or easy_install? [18:23:35] install [18:23:41] With what options? [18:24:55] setup.py install --root=${D} (where $D is the image directory of the pkg-manager) [18:26:11] What error is displayed? [18:26:25] ah, never mind... I see the problem. [18:26:36] yeah ? [18:26:42] Use "setup.py install --single-version-externally-managed --root=${D}" [18:26:57] Without that option, it's installing as an egg. [18:27:08] 0.6a11 will probably do that automatically if you set --root. [18:27:17] it MUST install it as egg - atm it uses data_dis wich breaks it as non-egg [18:27:30] aa_ is fixing it [18:28:00] yeah [18:28:05] almost fixed actually [18:28:10] :) [18:28:28] but before I used to steal the main AUTHORS and COPYING files from the distribution root, not sure what to do now [18:28:42] I kind of want to keep them there for old times sake [19:22:56] r0nny: done [19:23:07] also, amazingly now pythonsetup.py develop seems to wokr [19:23:36] but, it still wants to do something in site-packages, and I need to work out how to keep it local, for devs without root access on the box [19:42:30] ** whit has left IRC () [20:40:40] aa_, just use "setup.py develop -d /some/dir/on/PYTHONPATH" [20:41:45] pje: right, we made a run directory to install to and put it in pythonpath [20:41:54] all works beautifully, much nicer than before [20:42:13] I just mentioned that 'cause you said it was doing something in site-packages. [20:50:16] ** pje has left IRC ("Client exiting")