[00:53] whoops RAOF your edgers expired, renewed ya :) [00:55] Sarvatt: Thanks. I was going to ping you about that :) [01:13] Wooo!!! Internet is back. [13:27] RAOF, Sarvatt: I'm almost done packaging the new nvidia driver. Did you put the libraries back where they used to be in Mesa? [13:32] tseliot, did you integrate the new nvidia_supported script? [13:32] bjsnider: yep [13:33] can you post the relevant section of the rules file somewhere? [13:34] bjsnider: I'll commit my changes soon in this branch: http://github.com/tseliot/nvidia-graphics-drivers [13:34] k [13:57] bjsnider: ok, pushed (note: this is not final) [14:02] tseliot, where is the code that creates the /dev/nvidiactrl file? [14:02] bjsnider: ??? [14:03] bjsnider: you will only find packaging scripts there [14:06] well, i was looking for the correct settings for that file, i assume it is owned by root:video and that the user must be part of the video group [14:07] I wouldn't know [14:08] i have to get hold of someone who has the blob installed right now [14:08] unfortunately it got screwed up thanks to ext4 when i lost power [14:09] i always lose data with ext4 if the system shuts down unexpectedly [14:21] :-/ [14:55] tseliot: Re: the commit message, it’s in fact the README that is missing IDs (and has been for all releases of the driver). The README is only used by the new script to find the correct symbol from the blob. The resulting list is based on the blob, not the README. [14:57] tseliot: The script traverses through the .rodata symbols until it finds one that correlates to the README list closely enough. Then that list is converted to the modaliases format. [14:59] tseliot: The list from various versions of the blob has consistently had zero ID deletions (compared to README) and almost 100 % additions compared to the number of IDs in the README. [15:00] tseliot: So, currently the thresholds for picking the symbol are: deletions = 0, additions ≤ 1.5·n where n is the number of IDs in README. [15:00] ion: ah, it seemed weird to me that they removed only some ids. I'll read your changes again [15:01] tseliot: Sarvatt said the IDs README is missing are things like AGP-PCIe bridge chips etc. [15:02] ion: aah, that's where I heard that something was removed. I must have mixed things up [15:04] * tseliot changes the changelog [15:06] tseliot: For previous releases of the driver, the blob had two consecutive symbols duplicating the ID list, and that was the heuristic used to find the right symbol. They removed the duplicate and the script stopped working. So now the new heuristic for finding the right symbol is the comparison to the list from README. [15:13] tseliot: If you haven’t already done it, i could as well do the changelog change. [15:15] ion: I did this but I haven't pushed it yet: http://pastebin.ubuntu.com/455000/ [15:15] ion: if you want to add a detailed explanation to the changelog, you're welcome to do so [15:16] BTW why 1.5·n? [15:20] It’s pretty much an arbitrary number for sanity checking. The number of additions has correlated with the number of total IDs in README, and since it has consistently been *almost* 1.0·n, i added a safety buffer of about 50 % in case the number of additions exceeds the number of IDs in README in a future release. [15:20] ok, I suspected that but I wanted to be sure [15:21] ion: where's your branch on git-hub? [15:24] tseliot: Just pushed to http://github.com/ion1/nvidia-graphics-drivers / git://github.com/ion1/nvidia-graphics-drivers.git [15:26] ion: ok, I'll merge from it [15:33] merged and pushed. Thanks [15:47] the driver works well now [15:51] I guess nvidia-173 and nvidia-96 are not installable in maverick, are they? [15:51] bjsnider: ^ [15:54] Have you considered having just a single source package for all versions of the nvidia driver, btw? Wouldn’t that reduce duplication? I made sure the new nvidia_supported script works for all versions, for instance. It could just as well exist in a single source package. [15:55] ion, i added y our script to the older distros and it generated the expected modaliases file [15:56] ion: that would make the source package too big with all the installers [15:58] also bug tracking would be harder [15:58] Good point === virtuald_ is now known as virtuald