Mauro, I disagree with this change. It looks like you are attempting to workaround a Kbuild bug, by adding additional dependencies to modules that select FW_LOADER. Rather than doing this, we should work on fixing Kbuild such that this situation would be corrected. Energy would be better spent fixing the Kbuild issue in the kernel rather than working around the problem like this. Meanwhile, this looks wrong to me. These drivers are not necessarily hotplug drivers, but I understand that Hotplug is a dependency of FW_LOADER. It seems to me that the firmware loader code could (and should) be optimized such that it would no longer actually depend on hotplug. It's clear that there are two problems here. #1, Kbuild needs fixing. #2, FW_LOADER depends on Hotplug, but not all users of FW_LOADER are hotplug devices. Working around these problems are really masking the real issues. We should not mask these issues -- we should get them fixed upstream, instead. I would rather not see this patch applied to the kernel. Regards, Mike On Tue, May 6, 2008 at 10:10 AM, Patch from Mauro Carvalho Chehab --
On Tue, 6 May 2008 10:54:02 -0400
So, propose another way to fix. Without this patch, if HOTPLUG = 'n', several
drivers will break.
Sorry, but I can't see any other solution (in fact, I tested "select HOTPLUG",
but this caused circular dependencies loops).
The way "select" works is simple: You need to manually add all "depends on"
from the original symbol into the one that adds "select". I don't like this way
(and I'm currently questioning this at LKML), but while people don't decide to
implement "select" on another way (or to write a different tag for this, as
proposed by Sam), the proper fix is the one I've proposed.
If you have a better proposal, please, feel free to send it. If you have a
It is not a Kbuild bug, but the way Kbuild were implemented. This is what is
written at kbuild/kconfig-language.txt:
" select should be used with care. select will force
a symbol to a value without visiting the dependencies.
By abusing select you are able to select a symbol FOO even
if FOO depends on BAR that is not set.
In general use select only for non-visible symbols
(no prompts anywhere) and for symbols with no dependencies.
That will limit the usefulness but on the other hand avoid
the illegal configurations all over.
kconfig should one day warn about such things."
Cheers,
Mauro
--
| Yu Zhao | [PATCH 2/16 v6] PCI: define PCI resource names in an 'enum' |
| Greg Kroah-Hartman | [PATCH 011/196] sysfs: Fix a copy-n-paste typo in comment |
| Laurent Riffard | Re: 2.6.23-mm1: BUG in reiserfs_delete_xattrs |
| Ben Crowhurst | Kernel Development & Objective-C |
git: | |
| Kyle Rose | [OT] Re: C++ *for Git* |
| cte | linking libgit.a in C++ projects |
| Linus Torvalds | Re: CRLF problems with Git on Win32 |
| Pierre Habouzit | Re: [PATCH] bundle, fast-import: detect write failure |
| Thor Lancelot Simon | Re: sysctl knob to let sugid processes dump core (pr 15994) |
| YAMAMOTO Takashi | Re: Patches for EST and SMP |
| Bill Studenmund | Re: @booted_kernel magic symlink? |
| Adam Hamsik | Re: Thread benchmarks, round 2 |
| Chris | OpenBSD 4.4 installation error: write failed; file system full |
| Samuel Moñux | Cyrus IMAP performance problems [Long] |
| Steve B | Intel Atom and D945GCLF2 |
| James Hartley | scp batch mode? |
