On 12/18/18 3:09 PM, Joseph Myers wrote: >> diff --git a/sysdeps/arc/crti.S b/sysdeps/arc/crti.S > As a new port I think it would be best to use init_array in your Implies > file so you don't need these crti / crtn files, and make GCC generate > init_array / fini_array exclusively. (See RISC-V and C-Sky discussions of > this issue.) > I did the switch and while is working, there are additional test failures | FAIL: dlfcn/bug-atexit1 | FAIL: dlfcn/bug-atexit2 | FAIL: dlfcn/tstatexit | FAIL: dlfcn/tstcxaatexit | FAIL: nptl/tst-atfork2 | FAIL: nptl/tst-cond17 | FAIL: stdlib/test-dlclose-exit-race In the failing case, the dsos are NOT getting the DT_INIT/DT_FINI dynamic tags Consider glibc test dlfcn/tstatexit, which uses a dso modatexit.so In good case | arc-linux-readelf -a modatexit.os | | Section Headers: | [Nr] Name Type Addr Off Size ES Flg Lk Inf Al | [ 9] .init PROGBITS 0000047c 00047c 000044 00 AX 0 0 4 | [12] .fini PROGBITS 000006f8 0006f8 000028 00 AX 0 0 4 | | Dynamic section at offset 0x1f08 contains 23 entries: | 0x0000000c (INIT) 0x47c | 0x0000000d (FINI) 0x6f8 In broken case, init/fini sections exist, but DT_INI/DT_FINI are missing. | | Dynamic section at offset 0x1f2c contains 19 entries: <-- missing INIT/FINI | I suspected some bintuils/gcc shenanigans and tried hacking around, but was led far too astray. (1). gcc config test failing (false alaram) | host-gcc-final-ab544139bfee/build/gcc/config.log | |configure:22941: checking for .preinit_array/.init_array/.fini_array support |configure:23106: checking cross compile... guessing |configure:23111: result: no gcc was seemingly built w/o init_array support leading to inclusion of default gcc/config/initfini-array.h However doing a glibc aarch64 build I see the same output so perhaps it is not relevant. (2). ARC binutils port seems to hardwire _init/_fini (and the only arch to do that) so I ripped out the relevant handling from ld, but that doesn't help. https://github.com/foss-for-synopsys-dwc-arc-processors/binutils-gdb/commit/56771171fa7cb9e31c65e26a656a8588d603d924 So I'm not sure if this is deficiency of linker emulation script or code ld/bfd etc.. -Vineet _______________________________________________ linux-snps-arc mailing list linux-snps-arc@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/linux-snps-arc
