> Hi Thierry again Hello again, Andree > It's one call of sed and that's it, so that's cool. However running it, > we get e.g. for 'ldd /bin/grep'): > > libc.so.6 /lib/tls/libc.so.6 > /lib/ld-linux.so.2 /lib/ld-linux.so.2 > which means that things appear effectively twice for every row. Yes, I had let it be so because mindi sorts everything out anyway. But you're right, it can be made even cleaner.
> > The attached new patch mindi_patch_BTS324302_v5.diff makes it look like > this: > > libc.so.6 > /lib/ld-linux.so.2 > Also, the latest patch tries to precisely address the different > situations we may encounter: > > '[[:blank:]]*.*[[:blank:]]*=>[[:blank:]]*(.*/d' just get's rid of things > that are just in the kernel, i.e. point to a hex address. > > 's/[[:blank:]]*\(.*\)[[:blank:]]*=>[[:blank:]]*\/.*/\1/' deals with > normal library lines and ld-linux under old libc6. > > 's/[[:blank:]]*\(\/.*\)[[:blank:]]*(.*/\1/' handles ld-linux under new > libc6. > > What do you think? With this latest patch, we keep only the first part of ldd's output, so the libs in /tls don't get included. So it is OK for FAILSAFE and other 2.4 kernels. I tried with a 2.6 kernel, it doesn't seem to complain that there is no /lib/tls, and the mindi CD seems to work all right. But can we be really sure that we can suppress /lib/tls, without bad consequences, with any kernel, current or future ? 1 - If we can suppress /lib/tls : then we simply need to keep the first part of ldd's output, as this latest patch does, which can be made shorter with : sed 's/(.*)//; s/[[:blank:]]//g; /=>$/d; s/=>.*/ /' (tested to work in all cases) 2 - If we have to keep /lib/tls : then we can use the last patch I submitted - with one more command to avoid having twice the same lib : sed 's/(.*)//; s/[[:blank:]]//g; /=>$/d; s/=>/ /; s/\(..*\) \1/\1/' (tested as well) As I really don't know if option 1 is safe, I would keep 2. OTOH, if 1 is safe, then it should probably be prefered - more simple, and less libs means less space used on the small image... I leave it to you all who know better ! > Finally, could you repeat once again what the exact circumstances are > for which first patch doesn't work? I know, I'm a pain. It's good to have things put down clearly, please don't apologize ! :-) > The thing is > that the patch is actually suggested by other people as well and it is > relatively short and elegant. All good reasons to use it. But if it > doesn't work, we have to do something else... It fails when : - you run a 2.6 kernel and - you make a mindi CD with a FAILSAFE kernel So, the very thing I know now that one shouldn't do. But as long as it's possible (and people use it, as I innocently did two years long), I think that it should work and not leave you with a rescue CD that can't boot... But I completely agree with you, if it's something to avoid, it has to be made very clear ! Thierry -- Thierry Lathuille Annemasse, France -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]