This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [rfa] New test sigbpt.{c,exp}


mec> Can you make more of the test names more unique?
ac> Oops, I thought I'd covered that.  Try the attached.

It works for me.
All the test names are unique except for "rerun to main".
Stil the same four FAILs but that's okay because they really are FAILs.

I approve this patch.

ac> Are there [get_kernel_info] and [get_software_singlestep] testsuite calls?

I don't know of any.

I'm a bit leery of "get_kernel_info", because we should be testing
for features rather than version numbers (the whole autoconf philosophy).
Although get_compiler_info has worked reasonably well for the job
that it does.

I'm more inclined to dump a bunch of KFAIL's into the gdb_test_multiple
arms for the bad results, along with comments about which kernel versions
have been observed to have which behavior.  Then in 3-5 years it's not
hard to sort out the obsolete crap.

If KFAIL's are too hard then just the comments for starters.

I both kfailed and and commented, committing the attached.


thanks,
Andrew

2004-07-06  Andrew Cagney  <cagney@gnu.org>

	* gdb.base/sigbpt.exp, gdb.base/sigbpt.c: New test.

Index: gdb.base/sigbpt.c
===================================================================
RCS file: gdb.base/sigbpt.c
diff -N gdb.base/sigbpt.c
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ gdb.base/sigbpt.c	6 Jul 2004 15:08:19 -0000
@@ -0,0 +1,53 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+   Copyright 2004 Free Software Foundation, Inc.
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+ 
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+   02111-1307, USA.  */
+
+#include <signal.h>
+#include <stdlib.h>
+#include <string.h>
+
+extern void
+keeper (int sig)
+{
+}
+
+volatile long v1 = 0;
+volatile long v2 = 0;
+volatile long v3 = 0;
+
+extern long
+bowler (void)
+{
+  /* Try to read address zero.  Do it in a slightly convoluted way so
+     that more than one instruction is used.  */
+  return *(char *) (v1 + v2 + v3);
+}
+
+int
+main ()
+{
+  static volatile int i;
+
+  struct sigaction act;
+  memset (&act, 0, sizeof act);
+  act.sa_handler = keeper;
+  sigaction (SIGSEGV, &act, NULL);
+
+  bowler ();
+  return 0;
+}
Index: gdb.base/sigbpt.exp
===================================================================
RCS file: gdb.base/sigbpt.exp
diff -N gdb.base/sigbpt.exp
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ gdb.base/sigbpt.exp	6 Jul 2004 15:08:19 -0000
@@ -0,0 +1,263 @@
+# This testcase is part of GDB, the GNU debugger.
+
+# Copyright 2004 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.  
+
+# Check that GDB can and only executes single instructions when
+# stepping through a sequence of breakpoints interleaved by a signal
+# handler.
+
+# This test is known to tickle the following problems: kernel letting
+# the inferior execute both the system call, and the instruction
+# following, when single-stepping a system call; kernel failing to
+# propogate the single-step state when single-stepping the sigreturn
+# system call, instead resuming the inferior at full speed; GDB
+# doesn't know how to software single-step across a sigreturn
+# instruction.  Since the kernel problems can be "fixed" using
+# software single-step this is KFAILed rather than XFAILed.
+
+if $tracelevel {
+    strace $tracelevel
+}
+
+set prms_id 0
+set bug_id 0
+
+set testfile "sigbpt"
+set srcfile ${testfile}.c
+set binfile ${objdir}/${subdir}/${testfile}
+if { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable {debug}] != "" } {
+    gdb_suppress_entire_file "Testcase compile failed, so all tests in this file will automatically fail."
+}
+
+gdb_exit
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load ${binfile}
+
+#
+# Run to `main' where we begin our tests.
+#
+
+if ![runto_main] then {
+    gdb_suppress_tests
+}
+
+# If we can examine what's at memory address 0, it is possible that we
+# could also execute it.  This could probably make us run away,
+# executing random code, which could have all sorts of ill effects,
+# especially on targets without an MMU.  Don't run the tests in that
+# case.
+
+send_gdb "x 0\n"
+gdb_expect {
+    -re "0x0:.*Cannot access memory at address 0x0.*$gdb_prompt $" { }
+    -re "0x0:.*Error accessing memory address 0x0.*$gdb_prompt $" { }
+    -re ".*$gdb_prompt $" {
+	untested "Memory at address 0 is possibly executable"
+	return
+    }
+}
+
+gdb_test "break keeper"
+
+# Run to bowler, and then single step until there's a SIGSEGV.  Record
+# the address of each single-step instruction (up to and including the
+# instruction that causes the SIGSEGV) in bowler_addrs, and the address
+# of the actual SIGSEGV in segv_addr.
+
+set bowler_addrs bowler
+gdb_test {display/i $pc}
+gdb_test "advance *bowler" "bowler.*" "advance to the bowler"
+set test "stepping to SIGSEGV"
+gdb_test_multiple "stepi" "$test" {
+    -re "Program received signal SIGSEGV.*pc *(0x\[0-9a-f\]*).*$gdb_prompt $" {
+	set segv_addr $expect_out(1,string)
+	pass "$test"
+    }
+    -re " .*pc *(0x\[0-9a-f\]*).*bowler.*$gdb_prompt $" {
+	set bowler_addrs [concat $expect_out(1,string) $bowler_addrs]
+	send_gdb "stepi\n"
+	exp_continue
+    }
+}
+
+# Now record the address of the instruction following the faulting
+# instruction in bowler_addrs.
+
+set test "get insn after fault"
+gdb_test_multiple {x/2i $pc} "$test" {
+    -re "(0x\[0-9a-f\]*).*bowler.*(0x\[0-9a-f\]*).*bowler.*$gdb_prompt $" {
+	set bowler_addrs [concat $expect_out(2,string) $bowler_addrs]
+	pass "$test"
+    }
+}
+
+# Procedures for returning the address of the instruction before, at
+# and after, the faulting instruction.
+
+proc before_segv { } {
+    global bowler_addrs
+    return [lindex $bowler_addrs 2]
+}
+
+proc at_segv { } {
+    global bowler_addrs
+    return [lindex $bowler_addrs 1]
+}
+
+proc after_segv { } {
+    global bowler_addrs
+    return [lindex $bowler_addrs 0]
+}
+
+# Check that the address table and SIGSEGV correspond.
+
+set test "Verify that SIGSEGV occurs at the last STEPI insn"
+if {[string compare $segv_addr [at_segv]] == 0} {
+    pass "$test"
+} else {
+    fail "$test ($segv_addr [at_segv])"
+}
+
+# Check that the inferior is correctly single stepped all the way back
+# to a faulting instruction.
+
+proc stepi_out { name args } {
+    global gdb_prompt
+
+    # Set SIGSEGV to pass+nostop and then run the inferior all the way
+    # through to the signal handler.  With the handler is reached,
+    # disable SIGSEGV, ensuring that further signals stop the
+    # inferior.  Stops a SIGSEGV infinite loop when a broke system
+    # keeps re-executing the faulting instruction.
+    rerun_to_main
+    gdb_test "handle SIGSEGV nostop print pass" "" "pass SIGSEGV; $name"
+    gdb_test "continue" "keeper.*" "continue to keeper; $name"
+    gdb_test "handle SIGSEGV stop print nopass" "" "nopass SIGSEGV; $name"
+
+    # Insert all the breakpoints.  To avoid the need to step over
+    # these instructions, this is delayed until after the keeper has
+    # been reached.
+    for {set i 0} {$i < [llength $args]} {incr i} {
+	gdb_test "break [lindex $args $i]" "Breakpoint.*" \
+	    "set breakpoint $i of [llength $args]; $name"
+    }
+
+    # Single step our way out of the keeper, through the signal
+    # trampoline, and back to the instruction that faulted.
+    set test "stepi out of handler; $name"
+    gdb_test_multiple "stepi" "$test" {
+	-re "keeper.*$gdb_prompt $" {
+	    send_gdb "stepi\n"
+	    exp_continue
+	}
+	-re "signal handler.*$gdb_prompt $" {
+	    send_gdb "stepi\n"
+	    exp_continue
+	}
+	-re "Program received signal SIGSEGV.*$gdb_prompt $" {
+	    kfail gdb/1702 "$test (executed fault insn)"
+	}
+	-re "Breakpoint.*pc *[at_segv] .*bowler.*$gdb_prompt $" {
+	    pass "$test (at breakpoint)"
+	}
+	-re "Breakpoint.*pc *[after_segv] .*bowler.*$gdb_prompt $" {
+	    kfail gdb/1702 "$test (executed breakpoint)"
+	}
+	-re "pc *[at_segv] .*bowler.*$gdb_prompt $" {
+	    pass "$test"
+	}
+	-re "pc *[after_segv] .*bowler.*$gdb_prompt $" {
+	    kfail gdb/1702 "$test (skipped fault insn)"
+	}
+    }
+
+    # Clear any breakpoints
+    for {set i 0} {$i < [llength $args]} {incr i} {
+	gdb_test "clear [lindex $args $i]" "Deleted .*" \
+	    "clear breakpoint $i of [llength $args]; $name"
+    }
+}
+
+# Let a signal handler exit, returning to a breakpoint instruction
+# inserted at the original fault instruction.  Check that the
+# breakpoint is hit, and that single stepping off that breakpoint
+# executes the underlying fault instruction causing a SIGSEGV.
+
+proc cont_out { name args } {
+    global gdb_prompt
+
+    # Set SIGSEGV to pass+nostop and then run the inferior all the way
+    # through to the signal handler.  With the handler is reached,
+    # disable SIGSEGV, ensuring that further signals stop the
+    # inferior.  Stops a SIGSEGV infinite loop when a broke system
+    # keeps re-executing the faulting instruction.
+    rerun_to_main
+    gdb_test "handle SIGSEGV nostop print pass" "" "pass SIGSEGV; $name"
+    gdb_test "continue" "keeper.*" "continue to keeper; $name"
+    gdb_test "handle SIGSEGV stop print nopass" "" "nopass SIGSEGV; $name"
+
+    # Insert all the breakpoints.  To avoid the need to step over
+    # these instructions, this is delayed until after the keeper has
+    # been reached.  Always set a breakpoint at the signal trampoline
+    # instruction.
+    set args [concat $args "*[at_segv]"]
+    for {set i 0} {$i < [llength $args]} {incr i} {
+	gdb_test "break [lindex $args $i]" "Breakpoint.*" \
+	    "set breakpoint $i  of [llength $args]; $name"
+    }
+
+    # Let the handler return, it should "appear to hit" the breakpoint
+    # inserted at the faulting instruction.  Note that the breakpoint
+    # instruction wasn't executed, rather the inferior was SIGTRAPed
+    # with the PC at the breakpoint.
+    gdb_test "continue" "Breakpoint.*pc *[at_segv] .*" \
+	"continue to breakpoint at fault; $name"
+
+    # Now single step the faulted instrction at that breakpoint.
+    gdb_test "stepi" \
+	"Program received signal SIGSEGV.*pc *[at_segv] .*" \
+	"stepi fault; $name"    
+
+    # Clear any breakpoints
+    for {set i 0} {$i < [llength $args]} {incr i} {
+	gdb_test "clear [lindex $args $i]" "Deleted .*" \
+	    "clear breakpoint $i of [llength $args]; $name"
+    }
+
+}
+
+
+
+# Try to confuse DECR_PC_AFTER_BREAK architectures by scattering
+# breakpoints around the faulting address.  In all cases the inferior
+# should single-step out of the signal trampoline halting (but not
+# executing) the fault instruction.
+
+stepi_out "stepi"
+stepi_out "stepi bp before segv" "*[before_segv]"
+stepi_out "stepi bp at segv" "*[at_segv]"
+stepi_out "stepi bp before and at segv" "*[at_segv]" "*[before_segv]"
+
+
+# Try to confuse DECR_PC_AFTER_BREAK architectures by scattering
+# breakpoints around the faulting address.  In all cases the inferior
+# should exit the signal trampoline halting at the breakpoint that
+# replaced the fault instruction.
+cont_out "cont"
+cont_out "cont bp after segv" "*[before_segv]"
+cont_out "cont bp before and after segv" "*[before_segv]" "*[after_segv]"

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]