<Hagfish>hmm, you'd think that a team that did the work to deblob the kernel would also care about pregenerated headers, but maybe it hasn't had as much publicity as an issue yet <fossy>stikonas: needs some make feature we dont have <fossy>.21 might work, I just went with the newest lts <stikonas>Hagfish: generated headers are a known thing but I don't think it's considered that big of an issue in general by most people <stikonas>mostly because they are less opaque than binary blobs <stikonas>you can easily study how everything works even if you have generated files <stikonas>and they are probably more readable than lets say ./configure scripts <stikonas>although, if you have 200 MBs of them that's open for discusssion <stikonas>fossy: in the worst case we can also insert some intermediate version of make <stikonas>we mostly stayed with the old one because it just worked for us (until perl) <fossy>this seems to be what was used for gnewsense 1.0 <fossy>so now i just should be able to look at a .15 -> .16 changelog and check there's no new things <stikonas>hmm, this script seems much more crude than linux-libre but I guess we are not really using most of those drivers anyway <stikonas>almost all blobs should be in drivers/ directory <fossy>I'm just disabling all drivers + anything that looks suspicious added in 2.6.16 then this script <fossy>it's funny looking at this. almsot nothing in the changelog still is used today <stikonas>hopefully it will work with most of the modern drives <xentrac>in theory generated headers just factually document some interface you're trying to program to <xentrac>they can be wrong but otherwise it's hard for them to be malicious <fossy>lol this file ./fs/nls/nls_cp949.c <fossy>856K of autogenerated lookup tables <fossy>luckily i disabled NLS already <xentrac>ugh. that's what I need to be working on tonight actually :( <Hagfish>stikonas: thanks for the context about generated headers, that's really informative <fossy>/* Not sure what all the bytes in these char <fossy> * arrays do, but they're necessary to make <stikonas>at least those huge header files get rid of magic numbers <xentrac>(I mean wrong data you send to hardware can easily provoke security vulnerabilities) <fossy>gnewsense missed quite a bit, even in 2.6.15 <fossy>grep -E '^ *0x[0123456789ABCDEFabcdef]*, *' -r . is a very good command for finding these files <fossy>LMAO someone has made a binary blob for the boot logo of some device for all things <xentrac>well, how else do you expect them to represent the boot logo? As a Logo-style turtle graphics program? <Hagfish>or, i guess, put the SVG renderer in the toolchain and store the XML as the source <stikonas>fossy: hmm, maybe we should try newer kernel then? <stikonas>it should be possible to build newer gnu make earlier <fossy>stikonas: how would you feel abt merging linux 2.6.16 temporarily so that other work can continue and then i immediatley go onto moving to linux-libre - i'll remove as many pregened files as I can but with the knowledge there are still probably a few lurking around for maybe another week while i update the kernel ***rt is now known as robin