**** BEGIN LOGGING AT Sun Nov 09 03:00:00 2014 Nov 09 17:10:30 Intel developers add mandatory 'freeze' state w/o taking care of devices which do not have freeze_ops. Other developers aren't anyway better. Booted tosa to find that the MMC controller is br0ken. Also it looks like they broke suspend/resume in some cases. Nov 09 17:10:51 Damn careless people Nov 09 17:13:10 hi lumag Nov 09 17:13:15 Hello Nov 09 17:13:49 after all I solved yesterday...I was doing readlink before mounting... Nov 09 17:13:54 I'm really angry at some people. Nov 09 17:14:03 Argh. Nov 09 17:14:22 the code is still ugly... Nov 09 17:14:44 I'll send to Jay7 for refactoring ;) Nov 09 17:15:27 about freeze, does this impact the fill pxa? Nov 09 17:15:32 *full Nov 09 17:15:59 I don't know. I will check later. Nov 09 17:16:11 Most probably no - nobody except ACPI implements freeze_ops. Nov 09 17:20:33 PXA (esp. pxa27x) might have more idle states (think of cpu_idle driver). But adding freeze probably does not make sense. Nov 09 17:20:56 * lumag is going away - back in few hours. Maybe. Nov 09 17:24:33 bye Nov 09 19:40:07 hello ant__ Nov 09 20:16:27 wb Nov 09 20:21:29 lumag_, this is the concept (tested) Nov 09 20:21:31 http://pastebin.com/0MxrFG96 Nov 09 20:21:57 I think it can be done with a single if... t Nov 09 20:22:05 rying to clean it up Nov 09 20:22:54 absolute symlinks should not work - s is unitialized Nov 09 20:23:35 it does... Nov 09 20:25:18 Because s points to some random location in mem. Nov 09 20:26:15 ah, yes, lacks a malloc Nov 09 20:26:40 I could do s[150] Nov 09 20:28:15 No. Nov 09 20:29:29 After you return from function, your s will still point to stack. Nov 09 20:29:46 Next function will overwrite that area. Nov 09 20:29:56 next is..new kernel... Nov 09 20:30:05 kexec Nov 09 20:30:28 so it 'happens' to work ;) Nov 09 20:30:41 Currently yes. Who knows, what happens after somebody changes code in another place. Nov 09 20:30:45 :) Nov 09 20:31:33 I had found a nice example with the readlink in an if.. two liner checking the first chars of buf Nov 09 20:31:41 can't find it anymore... Nov 09 20:32:56 basically we only have to prepend MOUNTPOINT if the first readlink's buffer char exists and is '/' Nov 09 20:33:08 else do nothing Nov 09 20:34:28 smthg around http://pubs.opengroup.org/onlinepubs/009695399/functions/readlink.html Nov 09 20:36:35 http://linux.die.net/man/2/readlink Nov 09 20:36:53 That is with proper lstat, malloc, etc. Nov 09 20:37:40 After lstat I'd check if (!sb.st_mode & S_IFLNK) => not symlink Nov 09 20:37:52 Ah, even simpler Nov 09 20:38:25 if (!S_ISLNK(sb.st_mode) Nov 09 20:39:10 isn't it overkill? Nov 09 21:36:34 * lumag_ is a perfectionist Nov 09 22:23:10 it's fun to play with C now and then...I've never studied it properly, only high level languages / SQL Nov 09 22:52:01 ant__, and I'm a C guy. I know (some) Java, but no other strange languages. Nov 09 22:52:31 I had to spend much time on databases Nov 09 22:52:41 so vba, php, mysql Nov 09 22:54:14 I've found a much better solution fwiw Nov 09 22:54:32 (still hack but w/out mem leaks...) **** ENDING LOGGING AT Mon Nov 10 02:59:59 2014