Of course, this is a bug in the user-space compnent (libamu in this case), and sure, a code that crashes because it does not like the release string is a bad code, no argument about that. the idea of running a test to locate problematic component was not ment to be a responsibility of kernel-maintainer, it's only something that a distribution mitght do when cheme changes, to detect problems vefore crashes happen.
On Mon, Sep 9, 2013 at 5:57 PM, Chris Boot <[email protected]> wrote: > On 09/09/13 14:07, Yoram Bar Haim wrote: > > this work-around will make program think they run on top of 2.6.x > > kernel, which might be a bug if the program does have special behavior > > for 3.x.y > > With all due respect, it's a bug in a program if it crashes with a 3.x > uname string. If it then also misbehaves if it's given a 2.6.x string, > it's no more or less a bug in the program. In either case, the broken > program needs fixing. > > The option to present a 2.6.x uname was done mostly as a hack for 3rd > party software for which it can be difficult to obtain updates. There's > no excuse to use this hack for software that can be built from source. > > Personally, I have run into this issue with the 3ware/LSI RAID admin > tools, tw_cli and 3dm2, both of which crash unless there are 3 > dot-separated components in the kernel uname string. This annoys me > enough that I build my own kernel with the 3rd element added in (among > other tweaks), but I don't really expect Debian to go back to having > this. What would be welcome is a documented way of building your own > kernel that has this, as currently my hacks to do this are a bit ugly. > > I'm pretty sure software developers have been told not to attempt to > parse uname strings as they could be in any format, but that clearly > hasn't stopped anyone. It's still their problem, not the kernel's! :-) > > HTH, > Chris > > > On Mon, Sep 9, 2013 at 3:49 PM, Ben Hutchings <[email protected] > > <mailto:[email protected]>> wrote: > > > > On Mon, 2013-09-09 at 14:12 +0200, Yoram Bar Haim wrote: > > > using 3.x instead of 3.x.0 (when minor-minor version is 0) is done > by > > > kernel.org <http://kernel.org>, so I suppose it makes sense for > > Debian to follow this > > > scheme. > > > > Actually, the '.0' is still included if you build a kernel without > any > > stable updates and without Debian patches. But the '3.2.0' in wheezy > > sometimes confuses people into thinking we aren't applying any stable > > updates. I reckoned that userland in wheezy would be ready for 3.x > > releases and so it would be safe to drop the third part in jessie. > > > > By the way, the kernel has a workaround for programs that can't > handle > > 3.x.y releases, which also works for 3.x. Use setarch with the > > --uname-2.6 option. > > > > Ben. > > > > > it may be a good idea to run a test that logs all components > calling > > > uname, so they can be checked for version scheme bugs. > > > > > > > > > > > > On Sun, Sep 8, 2013 at 5:46 PM, Lior Kaplan <[email protected] > > <mailto:[email protected]>> wrote: > > > Hi, > > > > > > > > > > > > Just FYI, as you might be interested in > > > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=722145 > > > > > > > > > Kaplan > > > > > > > > > > > > > -- > > Ben Hutchings > > Man invented language to satisfy his deep need to complain. - Lily > > Tomlin > > > > > > -- > Chris Boot > [email protected] >
