Dear Joe Hershberger,
Hi Marek,
[...]
It's a RFC, lemme explain:
-$(obj)u-boot-spl.lds: $(LDSCRIPT) depend +$(obj)u-boot-spl.lst: $(LIBS)
$(OBJDUMP) -h $(LIBS) | \
Dump the section headers in all object files
sed -n -e 's/.*\(\.u_boot_list[^ ]\+\).*$$/\1/p' | \
Filter only .u_boot_list.* symbols
sed 's/\.[^\.]\+$$//' | \
Remove the last .[^.]$ from each line. That's because the last part behind . is the name of the variable, we don't need that for generation of subsection boundary symbols.
sed -n ':s /^.\+$$/ { p;s/^\(.*\)\.[^\.]*$$/\1/;b s }' | \
For each line in format .u_boot_list.x.y.z, generate a list of as such (list of substrings with . as separator): .u_boot_list.x.y.z .u_boot_list.x.y .u_boot_list.x .u_boot_list
sed -n 's/\./.#/g;h;s/$$/\a/p;g;s/$$/@/p;g;s/$$/~/p;' | \
For each line, print the line thrice, first time append \a at the end, second @ and third ~. Also replace each dot with .# . See [1] and look for the list under "The order_end Keyword" section.
LC_COLLATE=C sort -u | \
Sort the list, it's imperative to use the C collating here, because of [1]. This results in lines ending with \a float above all lines ending with either @ or continuing with further dots. Same goes for ~, which falls at the end.
sed 's/#//g' | \
Remove the hashes.
sed -n -e '/\a$$/ { s/\./_/g;s/\a$$/__start = .;/p; }'\-e '/~$$/ { s/\./_/g;s/~$$/__end = .;/p; }'\-e '/@$$/ { s/\(.*\)@$$/*(SORT(\1.*));/p }' >$@
Replace \a with __start ... so this defines the subsection start marker, ~ with __end for subsection end marker and lines ending with @ with the SORT(subsection.*) so they catch the contents of subsection.
Could you not reuse this complicated logic by defining a make function and then call + eval it? Something like this:
Can you elaborate more? I don't quite get it ... :-(
define list_rule_template $(1) : $(2) $(OBJDUMP) -h $^ | \ sed -n -e 's/.*(.u_boot_list[^ ]+).*$$/\1/p' | \ sed 's/.[^.]+$$//' | \ sed -n ':s /^.+$$/ { p;s/^(.*).[^.]*$$/\1/;b s }' | \ sed -n 's/./.#/g;h;s/$$/\a/p;g;s/$$/@/p;g;s/$$/~/p;' | \ LC_COLLATE=C sort -u | \ sed 's/#//g' | \ sed -n -e '/\a$$/ { s/./_/g;s/\a$$/__start = .;/p; }'\ -e '/~$$/ { s/./_/g;s/~$$/__end = .;/p; }'\ -e '/@$$/ { s/(.*)@$$/*(SORT(\1.*));/p }' >$@ endef ... $(eval $(call list_rule_template,$(obj)u-boot.lst,$(LIBBOARD) $(LIBS))) ... $(eval $(call list_rule_template,$(obj)u-boot-spl.lst,$(LIBS)))
+$(obj)u-boot-spl.lds: $(LDSCRIPT) $(obj)u-boot-spl.lst depend
$(CPP) $(CPPFLAGS) $(LDPPFLAGS) -ansi -D__ASSEMBLY__ -P - < $< > $@-Joe
[1] http://pubs.opengroup.org/onlinepubs/007908799/xbd/locale.html
Best regards, Marek Vasut