Bug 6732
Summary: | WARNING: /usr/local/src/linux-2.6.16/debian/linux-image-2.6.17/lib/modules/2.6.17/kernel/sound/drivers/opl4/snd-* needs unknown symbol snd_* | ||
---|---|---|---|
Product: | Drivers | Reporter: | Kseniya (akbara13) |
Component: | Sound(ALSA) | Assignee: | Jaroslav Kysela (perex) |
Status: | CLOSED PATCH_ALREADY_AVAILABLE | ||
Severity: | normal | CC: | bunk, cuse |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.17 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: | .config |
Description
Kseniya
2006-06-22 03:37:39 UTC
Please attach your .config . I can confirm this, so I'd appreciate if you reopen the report! I compiled all components as modules and the actual problem here is that snd_timer was not compiled as a module, causing these unresolved snd_timer_* symbols. As a quick and dirty workaround I modified sound/core/Kconfig from: ... config SND_TIMER tristate ... to: ... config SND_TIMER tristate "Foo timer" ... called 'make menuconfig' and configured "Foo timer" as module. After that all mandatory (PCM) drivers compiled and worked flawlessly. Now the unresolved OPL symbols is another story. The only driver I compile that has a dependency to the OPL stuff is the Adlib device driver. In this case snd-opl3-synth depends on snd-opl3-lib, where the first one was compiled as kernel module whereas the second one was not. I tried the same hack as I did to force compiling the snd-timer component as kernel module, but the problem was still persistent. Then I realized that non of my device drivers actually depends on snd-opl3-synth and simply dropped it from compilation for now. Please attach your .config . Created attachment 12103 [details]
.config
Argh, I got the root cause of these entire problems: I accidently selected CONFIG_SND_ADLIB=y instead of CONFIG_SND_ADLIB=m (the default when selecting in 'make *config'). The uploaded .config reproduces this issue. After setting CONFIG_SND_ADLIB=m everything went fine.
But the question is, shouldn't there be sanity checks for these kind of config mistakes? At least that mistake should IMO not affect the compilation of other sound device drivers, which it apparently does ATM.
Bug reproduced. Bug was fixed in kernel 2.6.19. |