[cdt-debug-dev] GDB-5.3

["Alain Magloire" <alain@xxxxxxx>]

FYI,

  gdb-5.3 exhibits some strange behaviour when dealing with the registers:

For example

gdb -i mi t.exe

(gdb)
-data-list-register-names
^done,register-names=["eax","ecx","edx","ebx","esp","ebp","esi","edi","eip","efl
ags","cs","ss","ds","es","fs","gs","st0","st1","st2","st3","st4","st5","st6","st
7","fctrl","fstat","ftag","fiseg","fioff","foseg","fooff","fop","mm0","mm1","mm2
","mm3","mm4","mm5","mm6","mm7"]
(gdb)
-data-list-register-names 35
^done,register-names=["mm3"]
(gdb)
-data-list-register-values x 35
^error,msg="bad register number"
(gdb)
-data-list-register-values x 35
^error,msg="bad register number"
(gdb)
-data-list-register-values x 39
^error,msg="bad register number"
(gdb)
-data-list-register-names 39
^done,register-names=["mm7"]


On GNU/Linux with gdb-5.3, doing this sort of thing twice will
crash/core gdb from an assert().


I will put workaround for this later in the head branch, since
we've agreed that gdb should work from 5.2.1 and up.

Thanks.