[-- Attachment #1: Type: text/plain, Size: 2292 bytes --] Hi... Here is a new patch from my side for insight. It adresses 2 issues. The small one first: In the tcl code of the stackwindow there is an instance variable maxwidth which is calculated but never used somewhere. My patch dares to remove it. The bigger one: I am frequently seeing output in insight on windows 64 bit: (Internal error: pc 0x0 in read in psymtab, but not in symtab.) Especially when debugging objectiveC code but can also see in sometimes when debugging plain c. This is triggered by stack unwinding (showing stack window in insight). I tracked it down to 2 problems. One of it is in insight itself and fixed with this patch. The C code of insight dealing with the backtrace generation was made somewhere in 2001 and remainded largely unchanged since then. It was initially "inspired" by the code directly in gdb. GDB has changed since then in that area. I updated insight's C code for generating the backtrace for the tcl part to match again gdb's code in stack.c. I could remove a bigger bunch of insight's private code in that area and replace it directly with a call to find_frame_funname() of gdb itself. I also removed some dead code here. From what I could see now insight no longer genarates this error message directly. I could still trigger it in some situations by just issuing a "bt" from either insight's console or from within native gdb. I have to dive into that. But with this patch insight itself should be clean now in this regard. I also updated copyright informations in the affected files. Changelog: 2012-04-02 Roland Schwingel <roland.schwingel@onevision.com> * generic/gdbtk-stack.c: Updated copyright informations. include "stack.h" and "solib.h" (when needed). (gdb_stack): Some reformatting. (get_frame_name): Updated usage of gdb api for backtrace generation. Some reformatting. Removed dead code. Also handle frame type ARCH_FRAME now. * library/stackwin.it[bh]: Remove dead instance variable maxwidth. Updated copyright informations. A general question: The direction of stack dump in insight is opposite to when issueing a backtrace inside gdb itself. Is this on purpose? Wouldn't it be nicer when insight's stack window would show the stack in the same order? Any comments? Is this ok? Roland [-- Attachment #2: stackwin.patch --] [-- Type: text/plain, Size: 7022 bytes --] diff -ruNp gdbtk_orig/generic/gdbtk-stack.c gdbtk/generic/gdbtk-stack.c --- gdbtk_orig/generic/gdbtk-stack.c 2012-03-30 09:29:15.000000000 +0200 +++ gdbtk/generic/gdbtk-stack.c 2012-04-02 16:13:11.351540500 +0200 @@ -1,5 +1,6 @@ /* Tcl/Tk command definitions for Insight - Stack. - Copyright (C) 2001-2012 Free Software Foundation, Inc. + Copyright (C) 2001, 2002, 2003, 2008, 2011 + Free Software Foundation, Inc. This file is part of GDB. @@ -26,6 +27,10 @@ #include "dictionary.h" #include "varobj.h" #include "arch-utils.h" +#include "stack.h" +#ifndef PC_SOLIB +#include "solib.h" +#endif #include <tcl.h> #include "gdbtk.h" @@ -475,24 +480,24 @@ gdb_stack (ClientData clientData, Tcl_In /* Find the outermost frame */ r = GDB_get_current_frame (&fi); if (r != GDB_OK) - return TCL_ERROR; + return TCL_ERROR; while (fi != NULL) { top = fi; - r = GDB_get_prev_frame (fi, &fi); - if (r != GDB_OK) - fi = NULL; + r = GDB_get_prev_frame (fi, &fi); + if (r != GDB_OK) + fi = NULL; } + result_ptr->obj_ptr = Tcl_NewListObj (0, NULL); + /* top now points to the top (outermost frame) of the stack, so point it to the requested start */ start = -start; - r = GDB_find_relative_frame (top, &start, &top); - - result_ptr->obj_ptr = Tcl_NewListObj (0, NULL); + r = GDB_find_relative_frame (top, &start, &top); if (r != GDB_OK) - return TCL_OK; + return TCL_OK; /* If start != 0, then we have asked to start outputting frames beyond the innermost stack frame */ @@ -503,8 +508,8 @@ gdb_stack (ClientData clientData, Tcl_In { get_frame_name (interp, result_ptr->obj_ptr, fi); r = GDB_get_next_frame (fi, &fi); - if (r != GDB_OK) - break; + if (r != GDB_OK) + break; } } } @@ -515,14 +520,14 @@ gdb_stack (ClientData clientData, Tcl_In /* A helper function for get_stack which adds information about * the stack frame FI to the caller's LIST. * - * This is stolen from print_frame_info in stack.c. + * This is stolen from print_frame_info/print_frame in stack.c. */ + static void get_frame_name (Tcl_Interp *interp, Tcl_Obj *list, struct frame_info *fi) { - struct symtab_and_line sal; struct symbol *func = NULL; - const char *funname = 0; + const char *funname = NULL; enum language funlang = language_unknown; Tcl_Obj *objv[1]; @@ -532,75 +537,39 @@ get_frame_name (Tcl_Interp *interp, Tcl_ Tcl_ListObjAppendElement (interp, list, objv[0]); return; } - if ((get_frame_type (fi) == SIGTRAMP_FRAME)) + if (get_frame_type (fi) == SIGTRAMP_FRAME) { objv[0] = Tcl_NewStringObj ("<signal handler called>", -1); Tcl_ListObjAppendElement (interp, list, objv[0]); return; } - - sal = - find_pc_line (get_frame_pc (fi), - get_next_frame (fi) != NULL - && !(get_frame_type (fi) == SIGTRAMP_FRAME) - && !(get_frame_type (fi) == DUMMY_FRAME)); - - func = find_pc_function (get_frame_pc (fi)); - if (func) - { - struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (get_frame_pc (fi)); - if (msymbol != NULL - && (SYMBOL_VALUE_ADDRESS (msymbol) - > BLOCK_START (SYMBOL_BLOCK_VALUE (func)))) - { - func = 0; - funname = GDBTK_SYMBOL_SOURCE_NAME (msymbol); - funlang = SYMBOL_LANGUAGE (msymbol); - } - else - { - funname = GDBTK_SYMBOL_SOURCE_NAME (func); - funlang = SYMBOL_LANGUAGE (func); - } - } - else + if (get_frame_type (fi) == ARCH_FRAME) { - struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (get_frame_pc (fi)); - if (msymbol != NULL) - { - funname = GDBTK_SYMBOL_SOURCE_NAME (msymbol); - funlang = SYMBOL_LANGUAGE (msymbol); - } + objv[0] = Tcl_NewStringObj ("<cross-architecture call>", -1); + Tcl_ListObjAppendElement (interp, list, objv[0]); + return; } - if (sal.symtab) + find_frame_funname (fi, &funname, &funlang, &func); + + if (funname) { objv[0] = Tcl_NewStringObj (funname, -1); Tcl_ListObjAppendElement (interp, list, objv[0]); } else { -#if 0 - /* we have no convenient way to deal with this yet... */ - if (fi->pc != sal.pc || !sal.symtab) - { - deprecated_print_address_numeric (fi->pc, 1, gdb_stdout); - printf_filtered (" in "); - } - printf_symbol_filtered (gdb_stdout, funname ? funname : "??", funlang, - DMGL_ANSI); -#endif - objv[0] = Tcl_NewStringObj (funname != NULL ? funname : "??", -1); + char *lib = NULL; + objv[0] = Tcl_NewStringObj (funname ? funname : "??", -1); #ifdef PC_SOLIB - if (!funname) - { - char *lib = PC_SOLIB (get_frame_pc (fi)); - if (lib) - { - Tcl_AppendStringsToObj (objv[0], " from ", lib, (char *) NULL); - } - } + lib = PC_SOLIB (get_frame_pc (fi)); +#else + lib = solib_name_from_address (get_frame_program_space (fi), + get_frame_pc (fi)); #endif + if (lib) + Tcl_AppendStringsToObj (objv[0], " from ", lib, (char *) NULL); + Tcl_ListObjAppendElement (interp, list, objv[0]); } } diff -ruNp gdbtk_orig/library/stackwin.itb gdbtk/library/stackwin.itb --- gdbtk_orig/library/stackwin.itb 2008-02-09 02:23:42.000000000 +0100 +++ gdbtk/library/stackwin.itb 2012-04-02 09:40:19.507691900 +0200 @@ -1,5 +1,5 @@ # Stack window for Insight. -# Copyright (C) 1997, 1998, 1999, 2002, 2003, 2008 Red Hat +# Copyright (C) 1997-2012 Red Hat, Inc. # # This program is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License (GPL) as published by @@ -68,20 +68,14 @@ itcl::body StackWin::update {event} { set frames {} } + $itk_component(slb) delete 0 end if {[llength $frames] == 0} { - $itk_component(slb) delete 0 end $itk_component(slb) insert end {NO STACK} return } - $itk_component(slb) delete 0 end set levels 0 foreach frame $frames { - set len [string length $frame] - - if {$len > $maxwidth} { - set maxwidth $len - } $itk_component(slb) insert end $frame incr levels } diff -ruNp gdbtk_orig/library/stackwin.ith gdbtk/library/stackwin.ith --- gdbtk_orig/library/stackwin.ith 2005-12-23 19:26:50.000000000 +0100 +++ gdbtk/library/stackwin.ith 2012-04-02 09:40:23.623824000 +0200 @@ -1,5 +1,5 @@ # Stack window class definition for GDBtk. -# Copyright (C) 1997, 1998, 1999 Cygnus Solutions +# Copyright (C) 1997-2012 Red Hat, Inc. # # This program is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License (GPL) as published by @@ -20,7 +20,6 @@ itcl::class StackWin { inherit EmbeddedWin GDBWin private { - variable maxwidth 40 variable Running 0 variable protect_me 0 method build_win {}
On 04/02/2012 08:21 AM, Roland Schwingel wrote: > > A general question: > The direction of stack dump in insight is opposite to when issueing a > backtrace inside gdb itself. Is this on purpose? Wouldn't it be nicer > when insight's stack window would show the stack in the same order? Yes, that was a design decision made many, many years ago. I don't have a preference to sort the stack "properly." If you find the time to write up a patch, I would consider/approve it. > diff -ruNp gdbtk_orig/generic/gdbtk-stack.c gdbtk/generic/gdbtk-stack.c > --- gdbtk_orig/generic/gdbtk-stack.c 2012-03-30 09:29:15.000000000 +0200 > +++ gdbtk/generic/gdbtk-stack.c 2012-04-02 16:13:11.351540500 +0200 > @@ -1,5 +1,6 @@ > /* Tcl/Tk command definitions for Insight - Stack. > - Copyright (C) 2001-2012 Free Software Foundation, Inc. > + Copyright (C) 2001, 2002, 2003, 2008, 2011 > + Free Software Foundation, Inc. > > This file is part of GDB. > You already did this. :-) > @@ -26,6 +27,10 @@ > #include "dictionary.h" > #include "varobj.h" > #include "arch-utils.h" > +#include "stack.h" > +#ifndef PC_SOLIB > +#include "solib.h" > +#endif > > #include<tcl.h> > #include "gdbtk.h" I believe the proper etiquette in gdb is to simply include solib.h unconditionally. It should not hurt anything. I tested this on mingw32 and x86_64-linux with no troubles. > @@ -475,24 +480,24 @@ gdb_stack (ClientData clientData, Tcl_In > /* Find the outermost frame */ > r = GDB_get_current_frame (&fi); > if (r != GDB_OK) > - return TCL_ERROR; > + return TCL_ERROR; > > while (fi != NULL) > { > top = fi; > - r = GDB_get_prev_frame (fi,&fi); > - if (r != GDB_OK) > - fi = NULL; > + r = GDB_get_prev_frame (fi,&fi); > + if (r != GDB_OK) > + fi = NULL; > } > > + result_ptr->obj_ptr = Tcl_NewListObj (0, NULL); > + > /* top now points to the top (outermost frame) of the > stack, so point it to the requested start */ > start = -start; > - r = GDB_find_relative_frame (top,&start,&top); > - > - result_ptr->obj_ptr = Tcl_NewListObj (0, NULL); > + r = GDB_find_relative_frame (top,&start,&top); Weird. My mail agent has removed the spaces after the commas. In the patch, all looks well. There does appear to be some whitespace trailing the "r = GDB_find_relative_frame (...);" line, though. Please double-check and remove if it is actually there. [and not another trick of my mail agent] > if (r != GDB_OK) > - return TCL_OK; > + return TCL_OK; > > /* If start != 0, then we have asked to start outputting > frames beyond the innermost stack frame */ > @@ -503,8 +508,8 @@ gdb_stack (ClientData clientData, Tcl_In > { > get_frame_name (interp, result_ptr->obj_ptr, fi); > r = GDB_get_next_frame (fi,&fi); > - if (r != GDB_OK) > - break; > + if (r != GDB_OK) > + break; "break;" doesn't appear indented properly, though. Please double-check. > } > } > } This rest of the whitespace change is okay, but please submit these sorts of patches separate from more substantial patches. It is a lot easier for me to review without that cluttering up the patch. > @@ -515,14 +520,14 @@ gdb_stack (ClientData clientData, Tcl_In > /* A helper function for get_stack which adds information about > * the stack frame FI to the caller's LIST. > * > - * This is stolen from print_frame_info in stack.c. > + * This is stolen from print_frame_info/print_frame in stack.c. > */ > + > static void > get_frame_name (Tcl_Interp *interp, Tcl_Obj *list, struct frame_info *fi) > { > - struct symtab_and_line sal; > struct symbol *func = NULL; > - const char *funname = 0; > + const char *funname = NULL; > enum language funlang = language_unknown; > Tcl_Obj *objv[1]; > > @@ -532,75 +537,39 @@ get_frame_name (Tcl_Interp *interp, Tcl_ > Tcl_ListObjAppendElement (interp, list, objv[0]); > return; > } > - if ((get_frame_type (fi) == SIGTRAMP_FRAME)) > + if (get_frame_type (fi) == SIGTRAMP_FRAME) > { > objv[0] = Tcl_NewStringObj ("<signal handler called>", -1); > Tcl_ListObjAppendElement (interp, list, objv[0]); > return; > } > - > - sal = > - find_pc_line (get_frame_pc (fi), > - get_next_frame (fi) != NULL > - && !(get_frame_type (fi) == SIGTRAMP_FRAME) > - && !(get_frame_type (fi) == DUMMY_FRAME)); > - > - func = find_pc_function (get_frame_pc (fi)); > - if (func) > - { > - struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (get_frame_pc (fi)); > - if (msymbol != NULL > - && (SYMBOL_VALUE_ADDRESS (msymbol) > - > BLOCK_START (SYMBOL_BLOCK_VALUE (func)))) > - { > - func = 0; > - funname = GDBTK_SYMBOL_SOURCE_NAME (msymbol); > - funlang = SYMBOL_LANGUAGE (msymbol); > - } > - else > - { > - funname = GDBTK_SYMBOL_SOURCE_NAME (func); > - funlang = SYMBOL_LANGUAGE (func); > - } > - } > - else > + if (get_frame_type (fi) == ARCH_FRAME) > { > - struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (get_frame_pc (fi)); > - if (msymbol != NULL) > - { > - funname = GDBTK_SYMBOL_SOURCE_NAME (msymbol); > - funlang = SYMBOL_LANGUAGE (msymbol); > - } > + objv[0] = Tcl_NewStringObj ("<cross-architecture call>", -1); > + Tcl_ListObjAppendElement (interp, list, objv[0]); > + return; > } > > - if (sal.symtab) > + find_frame_funname (fi,&funname,&funlang,&func); Argh! Darn mail agent! I apologize ahead of time if I ask you to watch out for this, and it turns out to be my mail agent and not you. > + > + if (funname) > { > objv[0] = Tcl_NewStringObj (funname, -1); > Tcl_ListObjAppendElement (interp, list, objv[0]); > } > else > { > -#if 0 > - /* we have no convenient way to deal with this yet... */ > - if (fi->pc != sal.pc || !sal.symtab) > - { > - deprecated_print_address_numeric (fi->pc, 1, gdb_stdout); > - printf_filtered (" in "); > - } > - printf_symbol_filtered (gdb_stdout, funname ? funname : "??", funlang, > - DMGL_ANSI); > -#endif > - objv[0] = Tcl_NewStringObj (funname != NULL ? funname : "??", -1); > + char *lib = NULL; Can you double-check the indent level on the above line? It doesn't look right. [Apologies again if my mail agent maligned it.] > + objv[0] = Tcl_NewStringObj (funname ? funname : "??", -1); > #ifdef PC_SOLIB > - if (!funname) > - { > - char *lib = PC_SOLIB (get_frame_pc (fi)); > - if (lib) > - { > - Tcl_AppendStringsToObj (objv[0], " from ", lib, (char *) NULL); > - } > - } > + lib = PC_SOLIB (get_frame_pc (fi)); > +#else > + lib = solib_name_from_address (get_frame_program_space (fi), > + get_frame_pc (fi)); GNU coding standards prefer that "get_frame_pc (fi));" be aligned underneath "get_frame_program_space (fi),". > #endif This has nothing to do with your patch other than the fact that you are touching this, but could you check the indentation on "#endif" while you're touching this code? It looks a little off. > + if (lib) > + Tcl_AppendStringsToObj (objv[0], " from ", lib, (char *) NULL); > + > Tcl_ListObjAppendElement (interp, list, objv[0]); > } > } > diff -ruNp gdbtk_orig/library/stackwin.itb gdbtk/library/stackwin.itb > --- gdbtk_orig/library/stackwin.itb 2008-02-09 02:23:42.000000000 +0100 > +++ gdbtk/library/stackwin.itb 2012-04-02 09:40:19.507691900 +0200 > @@ -1,5 +1,5 @@ > # Stack window for Insight. > -# Copyright (C) 1997, 1998, 1999, 2002, 2003, 2008 Red Hat > +# Copyright (C) 1997-2012 Red Hat, Inc. > # > # This program is free software; you can redistribute it and/or modify it > # under the terms of the GNU General Public License (GPL) as published by > @@ -68,20 +68,14 @@ itcl::body StackWin::update {event} { > set frames {} > } > > + $itk_component(slb) delete 0 end > if {[llength $frames] == 0} { > - $itk_component(slb) delete 0 end > $itk_component(slb) insert end {NO STACK} > return > } > > - $itk_component(slb) delete 0 end > set levels 0 > foreach frame $frames { > - set len [string length $frame] > - > - if {$len> $maxwidth} { > - set maxwidth $len > - } > $itk_component(slb) insert end $frame > incr levels > } > diff -ruNp gdbtk_orig/library/stackwin.ith gdbtk/library/stackwin.ith > --- gdbtk_orig/library/stackwin.ith 2005-12-23 19:26:50.000000000 +0100 > +++ gdbtk/library/stackwin.ith 2012-04-02 09:40:23.623824000 +0200 > @@ -1,5 +1,5 @@ > # Stack window class definition for GDBtk. > -# Copyright (C) 1997, 1998, 1999 Cygnus Solutions > +# Copyright (C) 1997-2012 Red Hat, Inc. > # > # This program is free software; you can redistribute it and/or modify it > # under the terms of the GNU General Public License (GPL) as published by > @@ -20,7 +20,6 @@ itcl::class StackWin { > inherit EmbeddedWin GDBWin > > private { > - variable maxwidth 40 > variable Running 0 > variable protect_me 0 > method build_win {} > Ok with those minor changes. Thanks! Keith
Hi... Thanks for reviewing Keith! I applied this now with all your comments taken into account. Some of the formatting things had also been on my side not only on your email clients side. Sorry for that! ATTENTION: At present insight is not compiling! This is not reasoned by my patch but due to changes inside of gdb itself. I will formulate a (trivial) patch for that. Roland