| 1 | |
| 2 | Known problems in GDB 6.2 |
| 3 | |
| 4 | See also: http://www.gnu.org/software/gdb/bugs/ |
| 5 | |
| 6 | |
| 7 | *** Build problems |
| 8 | |
| 9 | build/1458: compile failed on hpux11 |
| 10 | |
| 11 | GDB has build problems on HP/UX 11 with some versions of the HP |
| 12 | Ansi C compiler. (GCC works fine). |
| 13 | |
| 14 | The problem happens when compiling intl/bindtextdom.c. |
| 15 | The error is: |
| 16 | |
| 17 | cc: "gettextP.h", line 50: error 1000: Unexpected symbol: "SWAP". |
| 18 | cc: panic 2017: Cannot recover from earlier errors, terminating. |
| 19 | *** Error exit code 1 |
| 20 | |
| 21 | This is a problem with the 'inline' keyword in gettextP.h. |
| 22 | The workaround is to disable 'inline' before building gdb: |
| 23 | |
| 24 | export ac_cv_c_inline=no |
| 25 | |
| 26 | This problem happens only with some versions of the HP Ansi C compiler. |
| 27 | Versions A.11.01.25171.GP and B.11.11.28706.GP have both been observed |
| 28 | to work; version B.11.11.04 gets the build error and needs the |
| 29 | workaround. |
| 30 | |
| 31 | This problem might also happen with other C compilers. |
| 32 | |
| 33 | *** Misc |
| 34 | |
| 35 | gdb/1560: Control-C does not always interrupt GDB. |
| 36 | |
| 37 | When GDB is busy processing a command which takes a long time to |
| 38 | complete, hitting Control-C does not have the expected effect. |
| 39 | The command execution is not aborted, and the "QUIT" message confirming |
| 40 | the abortion is displayed only after the command has been completed. |
| 41 | |
| 42 | *** C++ support |
| 43 | |
| 44 | gdb/931: GDB could be more generous when reading types C++ templates on input |
| 45 | |
| 46 | When the user types a template, GDB frequently requires the type to be |
| 47 | typed in a certain way (e.g. "const char*" as opposed to "const char *" |
| 48 | or "char const *" or "char const*"). |
| 49 | |
| 50 | gdb/1512: no canonical way to output names of C++ types |
| 51 | |
| 52 | We currently don't have any canonical way to output names of C++ types. |
| 53 | E.g. "const char *" versus "char const *"; more subtleties arise when |
| 54 | dealing with templates. |
| 55 | |
| 56 | gdb/1516: [regression] local classes, gcc 2.95.3, dwarf-2 |
| 57 | |
| 58 | With gcc 2.95.3 and the dwarf-2 debugging format, classes which are |
| 59 | defined locally to a function include the demangled name of the function |
| 60 | as part of their name. For example, if a function "foobar" contains a |
| 61 | local class definition "Local", gdb will say that the name of the class |
| 62 | type is "foobar__Fi.0:Local". |
| 63 | |
| 64 | This applies only to classes where the class type is defined inside a |
| 65 | function, not to variables defined with types that are defined somewhere |
| 66 | outside any function (which most types are). |
| 67 | |
| 68 | gdb/1588: names of c++ nested types in casts must be enclosed in quotes |
| 69 | |
| 70 | You must type |
| 71 | (gdb) print ('Foo::Bar') x |
| 72 | or |
| 73 | (gdb) print ('Foo::Bar' *) y |
| 74 | instead of |
| 75 | (gdb) print (Foo::Bar) x |
| 76 | or |
| 77 | (gdb) print (Foo::Bar *) y |
| 78 | respectively. |
| 79 | |
| 80 | gdb/1091: Constructor breakpoints ignored |
| 81 | gdb/1193: g++ 3.3 creates multiple constructors: gdb 5.3 can't set breakpoints |
| 82 | |
| 83 | When gcc 3.x compiles a C++ constructor or C++ destructor, it generates |
| 84 | 2 or 3 different versions of the object code. These versions have |
| 85 | unique mangled names (they have to, in order for linking to work), but |
| 86 | they have identical source code names, which leads to a great deal of |
| 87 | confusion. Specifically, if you set a breakpoint in a constructor or a |
| 88 | destructor, gdb will put a breakpoint in one of the versions, but your |
| 89 | program may execute the other version. This makes it impossible to set |
| 90 | breakpoints reliably in constructors or destructors. |
| 91 | |
| 92 | gcc 3.x generates these multiple object code functions in order to |
| 93 | implement virtual base classes. gcc 2.x generated just one object code |
| 94 | function with a hidden parameter, but gcc 3.x conforms to a multi-vendor |
| 95 | ABI for C++ which requires multiple object code functions. |
| 96 | |
| 97 | *** Signal handlers |
| 98 | |
| 99 | On many systems an attempt to single-step a system-call instruction |
| 100 | results in two or more instructions being executed (the system-call, |
| 101 | and one or more instructions following). |
| 102 | |
| 103 | When attempting to single-step through a signal trampoline, this |
| 104 | problem may result the program unintentionally running to completion, |
| 105 | or re-execute the faulting instruction, or even corrupting the program |
| 106 | counter. |
| 107 | |
| 108 | Ref: PR breakpoints/1702. |
| 109 | |
| 110 | *** Stack backtraces |
| 111 | |
| 112 | GDB's core code base has been updated to use a new backtrace |
| 113 | mechanism. This mechanism makes it possible to support new features |
| 114 | such DWARF 2 Call Frame Information (which in turn makes possible |
| 115 | backtraces through optimized code). |
| 116 | |
| 117 | Since this code is new, it is known to still have a few problems: |
| 118 | |
| 119 | gdb/1505: [regression] gdb prints a bad backtrace for a thread |
| 120 | |
| 121 | When backtracing a thread, gdb does not stop when it reaches the |
| 122 | outermost frame, instead continuing until it hits garbage. This is |
| 123 | sensitive to the operating system and thread library. |