From: Thomas Petazzoni Date: Fri, 25 Sep 2020 09:45:46 +0200 Subject: [Buildroot] [PATCH 1/2] linux: show compression options only on some architectures In-Reply-To: <20200925073758.GD11621@scaer> References: <20200924122222.3449188-1-thomas.petazzoni@bootlin.com> <20200925073758.GD11621@scaer> Message-ID: <20200925094546.064d9c09@windsurf.hq.k.grp> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hello, On Fri, 25 Sep 2020 09:37:58 +0200 "Yann E. MORIN" wrote: > So, if the arch does not support kernel image compression, we will have > none of the compression symbols defined, which means that, in linux.mk, > we would have: > > 104 LINUX_COMPRESSION_OPT_$(BR2_LINUX_KERNEL_GZIP) += CONFIG_KERNEL_GZIP > 105 LINUX_COMPRESSION_OPT_$(BR2_LINUX_KERNEL_LZ4) += CONFIG_KERNEL_LZ4 > 106 LINUX_COMPRESSION_OPT_$(BR2_LINUX_KERNEL_LZMA) += CONFIG_KERNEL_LZMA > 107 LINUX_COMPRESSION_OPT_$(BR2_LINUX_KERNEL_LZO) += CONFIG_KERNEL_LZO > 108 LINUX_COMPRESSION_OPT_$(BR2_LINUX_KERNEL_XZ) += CONFIG_KERNEL_XZ > 109 LINUX_COMPRESSION_OPT_$(BR2_LINUX_KERNEL_ZSTD) += CONFIG_KERNEL_ZSTD > [--SNIP--] > 319 define LINUX_KCONFIG_FIXUP_CMDS > 322 $(call KCONFIG_ENABLE_OPT,$(strip $(LINUX_COMPRESSION_OPT_y))) > > But then, LINUX_COMPRESSION_OPT_y would be empty, so we'd call KCONFIG_ENABLE_OPT > with an empty string, which would expand to: > > $(call KCONFIG_MUNGE_DOT_CONFIG, $(1), $(1)=y, $(2)) > > ... and replacing the parameters would give (I'm pretty sure you'll see > where we're going now): > > $(call KCONFIG_MUNGE_DOT_CONFIG, , =y, $(2)) > > When KCONFIG_MUNGE_DOT_CONFIG is expanded, we'd get: > > $(SED) "/\\<$(strip $(1))\\>/d" $(call KCONFIG_DOT_CONFIG,$(3)) > echo '$(strip $(2))' >> $(call KCONFIG_DOT_CONFIG,$(3)) > > ... which would eventualy lead us to: > > $(SED) "/\\<)\\>/d" /some/.config-file > echo '=y' >> /some/.config-file > > And to be honest, I am not sure that last line would generate a fully > legit kconfig setting... ;-) > > Of course, running 'make oldconfig' on that modified .config would drop > invalid lines, so you would not notivce the issue, but still this is not > correct... Thanks for spotting this, I'll fix in v2. That being said, one can notice that you're pointing out that this bogus =y line will be dropped... just as equally as the bogus CONFIG_KERNEL_GZIP=y that we have today gets dropped on arm64 and other architectures for which those CONFIG_KERNEL_{GZIP,LZMA,LZ4,...} options don't exist :-) Thomas -- Thomas Petazzoni, CTO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com