1- old GNU ld's behavior wrt DSOs seems to be severely broken. 2 3 y.o reference foo() 4 y1.o defines foo(), references bar() 5 y2.o defines bar() 6 libbar.so defines bar() 7 8 Running 9 10 gcc -o y y.o -lbar y1.o y2.o 11 12 uses the bar() definition from libbar.so and does not mention the definition 13 in y2.o at all (no duplicate symbol message). Correct is to use the 14 definition in y2.o. 15 16 17 y.o reference foo() 18 y1.o defines foo(), references bar() 19 y2.o in liby2.a defines bar() 20 libbar.so defines bar() 21 22 Running 23 24 gcc -o y y.o -lbar y1.o -ly3 25 26 has to use the definition in -lbar and not pull the definition from liby3.a. 27 28 29- the old linker follows DT_NEEDED entries and adds the objects referenced 30 this way which define a symbol which is needed as a DT_NEEDED to the 31 generated binary. This is wrong since the DT_NEEDED changes the search 32 path in the object (which is breadth first). 33 34 35- the old linker supported extern "C++", extern "java" in version scripts. 36 I believe this implementation is severly broken and needs a redesign 37 (how do wildcards work with these languages*?). Therefore it is left 38 out for now. 39 40 41- what should happen if two sections in different files with the same 42 name have different types and/or the flags are different 43 44 45- section names in input files are mostly irrelevant. Exceptions: 46 47 .comment/SHT_PROGBITS in strip, ld 48 49 .debug \ 50 .line | 51 .debug_srcinfo | 52 .debug_sfnames | 53 .debug_aranges | 54 .debug_pubnames | 55 .debug_info | 56 .debug_abbrev | 57 .debug_line | 58 .debug_abbrev > DWARF sections in ld 59 .debug_line | 60 .debug_frame | 61 .debug_str | 62 .debug_loc | 63 .debug_macinfo | 64 .debug_weaknames | 65 .debug_funcnames | 66 .debug_typenames | 67 .debug_varnames / 68 69 Sections created in output files follow the naming of special section 70 from the gABI. 71 72 In no place is a section solely indentified by its name. Internal 73 references always use the section index. 74