* David Miller <davem@davemloft.net> wrote:i think any build which has the media drivers only as modules will trigger it, so it's rather widespread. My fix has held up fine for 10 randconfigs so far so it should do the trick. i'm wondering though why similar problems are not more common. A quick look at drivers/Makefile suggests that there are many obj-y rules there for driver sub-trees that can be all-modular, not just video: obj-y += video/ obj-y += char/ obj-y += serial/ obj-y += base/ block/ misc/ mfd/ net/ media/ obj-y += macintosh/ obj-y += ieee1394/ obj-y += cdrom/ obj-y += auxdisplay/ obj-y += i2c/ obj-y += lguest/ obj-y += firmware/ drivers/net/ obj-y is kept alive by the "obj-y wireless/" rule i think, but what makes other sub-trees link, like drivers/macintosh/ ? so kbuild apparently has smarts to handle this case - but somehow it still failed for drivers/media/. Sam Cc:-ed - am i missing some detail in this analysis? hm, one weird rule is drivers/video/matrox/Makefile's my-obj- rule. Ingo --
| Vladislav Bolkhovitin | Re: Integration of SCST in the mainstream Linux kernel |
| Greg Kroah-Hartman | [PATCH 001/196] Chinese: Add the known_regression URI to the HOWTO |
| Tarkan Erimer | Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 |
| Martin Michlmayr | Network slowdown due to CFS |
git: | |
| Paweł Staszewski | rib_trie / Fix inflate_threshold_root. Now=15 size=11 bits |
| David Miller | [GIT]: Networking |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| David Miller | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
