#milkymist IRC log for Tuesday, 2013-10-08

ysionneauin a linker script, when you declare memory regions with the "MEMORY" keyword, the ORIGIN= parameter indicates the LMA, right?06:14
ysionneaumeans physical address06:15
ysionneauI've got the error : /Users/fallen/dev/NetBSD/obj/tooldir.Darwin-10.8.0-i386/bin/lm32--netbsd-ld: address 0xc00eae1c of netbsd section `.text' is not within region `sdram'06:16
ysionneaubecause I put : MEMORY { sdram : ORIGIN = 0x40000000, LENGTH = 0x8000000 } SECTIONS { .text 0xc0000000 : AT ( 0x40000000 ) { .... } }06:17
ysionneauerr, I mean I put : MEMORY { sdram : ORIGIN = 0x40000000, LENGTH = 0x8000000 } SECTIONS { .text 0xc0000000 : AT ( 0x40000000 ) { .... } > sdram }06:18
ysionneaubut the 0xc00000 next to the section name is supposed to be a VMA ... not a LMA06:19
ysionneauif I remove completely the notion of MEMORY (and the "> sdram") it almost links, it just complains about a mysterious error: no memory region specified for loadable section `link_set_bufq_strats'06:25
ysionneauwhich I cannot find in the entire source code (or .o files)06:26
ysionneaugotta leave for work06:29
ysionneauif someone wants to have a look at my linker issue: http://pastebin.com/Ukp0q4qV20:54
ysionneauI posted the error, the linker script I used and a few command outputs20:54
ysionneauI basically end up with a section that I don't know where it comes from20:55
ysionneauand it makes the link fail...20:55
ysionneauso this kind of symbol might come from macro generated stuff21:36
ysionneaubut it's not just symbols, it's a section as well21:36
ysionneauah, got it21:38
ysionneau#define __link_set_make_entry(set, sym)                                 \21:38
ysionneau        static void const * const __link_set_##set##_sym_##sym          \21:38
ysionneau            __section("link_set_" #set) __used = (const void *)&sym21:38
ysionneauthis kind of stuff is used a bit in the kernel21:38
ysionneauit generates sections...21:38
ysionneauso now I put the link_set_* sections with the .rodata stuff21:43
ysionneaunow it does not complain anymore about this kind of sections and overlaps21:43
ysionneaubut it complains about a whole bunch of undefined symbols21:43
ysionneaulike __start_link_set_dkwedge_methods and __stop_link_set_dkwedge_methods21:44
ysionneauI guess it's just markers for dynamic regions21:45
ysionneaubut weird that nothing defines their addresses21:45
Action: ysionneau calling it a day21:45
--- Wed Oct 9 201300:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!