https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71776
--- Comment #8 from malithyapa at gmail dot com --- I managed to work around this by building gcc with --with-mode=arm Configured with: ../gcc-4.9.4/configure --enable-languages=c,c++ --with-mode=arm --with-float=hard --build=arm-linux-gnueabihf --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch=armv7-a --with-tune=cortex-a7 --with-sysroot=/ --libexecdir=/usr/lib --libdir=/usr/lib --enable-threads=posix but when i look at the thumb code, the linker should not have any problem there because the symbol is very much in range of the branch instruction (12 bit two's complement offset) 000000c4 <__gnu_h2f_internal>: c4: f3c0 2384 ubfx r3, r0, #10, #5 c8: 2900 cmp r1, #0 ca: f1a3 021f sub.w r2, r3, #31 ce: fab2 f282 clz r2, r2 d2: ea4f 1252 mov.w r2, r2, lsr #5 d6: bf08 it eq d8: 2200 moveq r2, #0 da: 4611 mov r1, r2 dc: f400 4200 and.w r2, r0, #32768 ; 0x8000 e0: f3c0 0009 ubfx r0, r0, #0, #10 e4: 0412 lsls r2, r2, #16 e6: b969 cbnz r1, 104 <__gnu_h2f_internal+0x40> e8: b92b cbnz r3, f6 <__gnu_h2f_internal+0x32> ea: b180 cbz r0, 10e <__gnu_h2f_internal+0x4a> ec: fab0 f380 clz r3, r0 f0: 3b15 subs r3, #21 f2: 4098 lsls r0, r3 f4: 425b negs r3, r3 f6: 3370 adds r3, #112 ; 0x70 f8: 0340 lsls r0, r0, #13 fa: eb00 53c3 add.w r3, r0, r3, lsl #23 fe: ea43 0002 orr.w r0, r3, r2 102: 4770 bx lr 104: ea42 3040 orr.w r0, r2, r0, lsl #13 108: f040 40ff orr.w r0, r0, #2139095040 ; 0x7f800000 10c: 4770 bx lr 10e: 4610 mov r0, r2 110: 4770 bx lr 112: bf00 nop 00000114 <__gnu_f2h_ieee>: 114: 2101 movs r1, #1 116: e773 b.n 0 <__gnu_f2h_internal> 00000118 <__gnu_h2f_ieee>: 118: 2101 movs r1, #1 11a: e7fe b.n c4 <__gnu_h2f_internal> ***!!!!!!!!!!!!!!*** The only reason i can think of is if LTO is enabled and it has done something funny with the organization, but -O0 was specifically mentioned in compilation. I don't understand the internals of gcc well enough to debug this, but the more i look the more it looks like a bug.
