From: Pedro Alves <palves@redhat.com>
To: gdb-patches@sourceware.org
Subject: [PATCH 36/36] Make TRY/CATCH use real C++ try/catch in C++ mode
Date: Mon, 09 Feb 2015 23:22:00 -0000 [thread overview]
Message-ID: <1423524046-20605-37-git-send-email-palves@redhat.com> (raw)
In-Reply-To: <1423524046-20605-1-git-send-email-palves@redhat.com>
Although the current TRY/CATCH implementation works in C++ mode too,
it relies on setjmp/longjmp, and longjmp bypasses calling the
destructors of objects on the stack, which is obviously bad for C++.
This patch fixes this by makes TRY/CATCH use real try/catch in C++
mode behind the scenes. The way this is done allows RAII and cleanups
to coexist while we phase out cleanups, instead of requiring a flag
day.
This patch is not strictly necessary until we require a C++ compiler
and start actually using RAII, though I'm all for baby steps, and it
shows my proposed way forward. I think we should put it in now, to
allow easier experimentation and exposure of potential problems with
real C++ exceptions.
gdb/ChangeLog:
2015-02-09 Pedro Alves <palves@redhat.com>
* common/common-exceptions.c [!__cplusplus] (enum catcher_state)
(exceptions_state_mc_action_iter)
(exceptions_state_mc_action_iter_1, exceptions_state_mc_catch):
Don't define.
[__cplusplus] (try_scope_depth): New global.
[__cplusplus] (exception_try_scope_entry)
(exception_try_scope_exit, gdb_exception_sliced_copy)
(exception_rethrow): New functions.
(throw_exception): In C++ mode, throw
gdb_exception_RETURN_MASK_QUIT for RETURN_QUIT and
gdb_exception_RETURN_MASK_ERROR for RETURN_ERROR.
(throw_it): In C++ mode, use try_scope_depth.
* common/common-exceptions.h [!__cplusplus]
(exceptions_state_mc_action_iter)
(exceptions_state_mc_action_iter_1, exceptions_state_mc_catch):
Don't declare.
[__cplusplus] (exception_try_scope_entry)
(exception_try_scope_exit, exception_rethrow): Declare.
[__cplusplus] (struct exception_try_scope): New struct.
[__cplusplus] (TRY, CATCH, END_CATCH): Reimplement on top of real
C++ exceptions.
(struct gdb_exception_RETURN_MASK_ALL)
(struct gdb_exception_RETURN_MASK_ERROR)
(struct gdb_exception_RETURN_MASK_QUIT): New types.
---
gdb/common/common-exceptions.c | 77 +++++++++++++++++++++++++++++++++++++++++-
gdb/common/common-exceptions.h | 67 ++++++++++++++++++++++++++++++++++++
2 files changed, 143 insertions(+), 1 deletion(-)
diff --git a/gdb/common/common-exceptions.c b/gdb/common/common-exceptions.c
index 2ad0ce1..b300a9d 100644
--- a/gdb/common/common-exceptions.c
+++ b/gdb/common/common-exceptions.c
@@ -22,6 +22,8 @@
const struct gdb_exception exception_none = { 0, GDB_NO_ERROR, NULL };
+#ifndef __cplusplus
+
/* Possible catcher states. */
enum catcher_state {
/* Initial state, a new catcher has just been created. */
@@ -185,7 +187,7 @@ exceptions_state_mc_catch (struct gdb_exception *exception,
{
if (mask & RETURN_MASK (exception->reason))
{
- /* Exit normally and let the called handle the
+ /* Exit normally and let the caller handle the
exception. */
return 1;
}
@@ -211,6 +213,56 @@ exceptions_state_mc_action_iter_1 (void)
return exceptions_state_mc (CATCH_ITER_1);
}
+#else /* !__cplusplus */
+
+/* How many nested TRY blocks we have. See exception_messages and
+ throw_it. */
+
+static int try_scope_depth;
+
+/* Called on entry to a TRY scope. */
+
+void *
+exception_try_scope_entry (void)
+{
+ ++try_scope_depth;
+ return (void *) save_cleanups ();
+}
+
+/* Called on exit of a TRY scope, either normal exit or exception
+ exit. */
+
+void
+exception_try_scope_exit (void *saved_state)
+{
+ restore_cleanups ((struct cleanup *) saved_state);
+ --try_scope_depth;
+}
+
+/* Called by the default catch block. IOW, we'll get here before
+ jumping out to the next outermost scope an exception if a GDB
+ exception is not caught. */
+
+void
+exception_rethrow (void)
+{
+ /* Run this scope's cleanups before re-throwing to the next
+ outermost scope. */
+ prepare_to_throw_exception ();
+ do_cleanups (all_cleanups ());
+ throw;
+}
+
+/* Copy the 'gdb_exception' portion of FROM to TO. */
+
+static void
+gdb_exception_sliced_copy (struct gdb_exception *to, const struct gdb_exception *from)
+{
+ *to = *from;
+}
+
+#endif /* !__cplusplus */
+
/* Return EXCEPTION to the nearest containing catch_errors(). */
void
@@ -220,12 +272,31 @@ throw_exception (struct gdb_exception exception)
do_cleanups (all_cleanups ());
+#ifndef __cplusplus
/* Jump to the containing catch_errors() call, communicating REASON
to that call via setjmp's return value. Note that REASON can't
be zero, by definition in defs.h. */
exceptions_state_mc (CATCH_THROWING);
current_catcher->exception = exception;
SIGLONGJMP (current_catcher->buf, exception.reason);
+#else
+ if (exception.reason == RETURN_QUIT)
+ {
+ gdb_exception_RETURN_MASK_QUIT ex;
+
+ gdb_exception_sliced_copy (&ex, &exception);
+ throw ex;
+ }
+ else if (exception.reason == RETURN_ERROR)
+ {
+ gdb_exception_RETURN_MASK_ERROR ex;
+
+ gdb_exception_sliced_copy (&ex, &exception);
+ throw ex;
+ }
+ else
+ gdb_assert_not_reached ("invalid return reason");
+#endif
}
/* A stack of exception messages.
@@ -249,7 +320,11 @@ throw_it (enum return_reason reason, enum errors error, const char *fmt,
{
struct gdb_exception e;
char *new_message;
+#ifndef __cplusplus
int depth = catcher_list_size ();
+#else
+ int depth = try_scope_depth;
+#endif
gdb_assert (depth > 0);
diff --git a/gdb/common/common-exceptions.h b/gdb/common/common-exceptions.h
index d2c0bee..2e6a6d9 100644
--- a/gdb/common/common-exceptions.h
+++ b/gdb/common/common-exceptions.h
@@ -121,10 +121,16 @@ struct gdb_exception
the exceptions subsystem and not used other than via the TRY/CATCH
macros defined below. */
+#ifndef __cplusplus
extern SIGJMP_BUF *exceptions_state_mc_init (void);
extern int exceptions_state_mc_action_iter (void);
extern int exceptions_state_mc_action_iter_1 (void);
extern int exceptions_state_mc_catch (struct gdb_exception *, int);
+#else
+extern void *exception_try_scope_entry (void);
+extern void exception_try_scope_exit (void *saved_state);
+extern void exception_rethrow (void);
+#endif
/* Macro to wrap up standard try/catch behavior.
@@ -151,6 +157,8 @@ extern int exceptions_state_mc_catch (struct gdb_exception *, int);
*/
+#ifndef __cplusplus
+
#define TRY \
{ \
SIGJMP_BUF *buf = \
@@ -168,6 +176,65 @@ extern int exceptions_state_mc_catch (struct gdb_exception *, int);
#define END_CATCH \
}
+#else
+
+/* Prevent error/quit during TRY from calling cleanups established
+ prior to here. This pops out the scope in either case of normal
+ exit or exception exit. */
+struct exception_try_scope
+{
+ exception_try_scope ()
+ {
+ saved_state = exception_try_scope_entry ();
+ }
+ ~exception_try_scope ()
+ {
+ exception_try_scope_exit (saved_state);
+ }
+
+ void *saved_state;
+};
+
+/* We still need to wrap TRY/CATCH in C++ so that cleanups and C++
+ exceptions can coexist. The TRY blocked is wrapped in a
+ do/while(0) so that break/continue within the block works the same
+ as in C. */
+#define TRY \
+ try \
+ { \
+ exception_try_scope exception_try_scope_instance; \
+ do \
+ {
+
+#define CATCH(EXCEPTION, MASK) \
+ } while (0); \
+ } \
+ catch (struct gdb_exception ## _ ## MASK &EXCEPTION)
+
+#define END_CATCH \
+ catch (...) \
+ { \
+ exception_rethrow (); \
+ }
+
+/* The exception types client code may catch. They're just shims
+ around gdb_exception that add nothing but type info. Which is used
+ is selected depending on the MASK argument passed to CATCH. */
+
+struct gdb_exception_RETURN_MASK_ALL : public gdb_exception
+{
+};
+
+struct gdb_exception_RETURN_MASK_ERROR : public gdb_exception_RETURN_MASK_ALL
+{
+};
+
+struct gdb_exception_RETURN_MASK_QUIT : public gdb_exception_RETURN_MASK_ALL
+{
+};
+
+#endif
+
/* *INDENT-ON* */
/* Hook to allow client-specific actions to be performed prior to
--
1.9.3
next prev parent reply other threads:[~2015-02-09 23:22 UTC|newest]
Thread overview: 100+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-02-09 23:20 [PATCH 00/36] Support building GDB as a C++ program Pedro Alves
2015-02-09 23:20 ` [PATCH 02/36] Add --enable-build-with-cxx configure switch Pedro Alves
2015-02-10 22:11 ` Yao Qi
2015-02-27 13:24 ` Pedro Alves
2015-02-27 14:20 ` Yao Qi
2015-02-27 16:29 ` Pedro Alves
2015-02-09 23:21 ` [PATCH 11/36] Make functions and variables exported by the IPA be extern "C" Pedro Alves
2015-02-09 23:21 ` [PATCH 17/36] mi/mi-cmd-stack.c|frame filters: print_values <-> ext_lang_frame_args Pedro Alves
2015-02-09 23:21 ` [PATCH 25/36] python/python-internal.h: enum ‘ext_lang_rc’ not defined Pedro Alves
2015-02-09 23:21 ` [PATCH 28/36] Move exception_none to common code, and use it Pedro Alves
2015-02-09 23:21 ` [PATCH 19/36] Exported const objects Pedro Alves
2015-02-09 23:21 ` [PATCH 16/36] x86 Linux/ptrace: fix offsetof usage in C++ mode Pedro Alves
2015-02-09 23:21 ` [PATCH 03/36] C++ keyword cleanliness, mostly auto-generated Pedro Alves
2015-02-11 7:57 ` Joel Brobecker
2015-02-11 8:52 ` Phil Muldoon
2015-02-11 10:27 ` Pedro Alves
2015-02-11 10:51 ` Pedro Alves
2015-02-12 12:19 ` Joel Brobecker
2015-02-12 13:14 ` Pedro Alves
2015-02-12 14:43 ` Pedro Alves
2015-02-12 14:59 ` Joel Brobecker
2015-02-27 17:41 ` Pedro Alves
2015-02-09 23:21 ` [PATCH 26/36] Adjust self tests to cope with GDB built as a C++ program Pedro Alves
2015-02-09 23:21 ` [PATCH 04/36] Fix struct, union, and enum nesting in C++ Pedro Alves
2015-02-09 23:21 ` [PATCH 27/36] catch_command_errors: Remove 'mask' parameter Pedro Alves
2015-02-09 23:21 ` [PATCH 29/36] Normalize TRY_CATCH exception handling block Pedro Alves
2015-02-09 23:21 ` [PATCH 21/36] opcodes/microblaze: Rename 'or', 'and', 'xor' to avoid C++ conflict Pedro Alves
2015-02-10 15:05 ` Michael Eager
2015-02-10 18:11 ` Pedro Alves
2015-02-09 23:21 ` [PATCH 13/36] target.h: Include infrun.h Pedro Alves
2015-02-09 23:21 ` [PATCH 20/36] gdbserver/tracepoint: Add cast sockaddr_un/sockaddr cast Pedro Alves
2015-02-09 23:21 ` [PATCH 08/36] elf-bfd.h: Wrap in extern "C" Pedro Alves
2015-02-09 23:33 ` Andrew Pinski
2015-02-10 12:05 ` [PATCH v2] Wrap BFD headers " Pedro Alves
2015-02-11 0:36 ` Alan Modra
2015-02-11 10:08 ` Pedro Alves
2015-02-09 23:21 ` [PATCH 09/36] floatformat.h: Wrap " Pedro Alves
2015-02-09 23:35 ` Andrew Pinski
2015-02-09 23:49 ` Pedro Alves
2015-02-12 11:49 ` Pedro Alves
2015-02-18 19:55 ` Jakub Jelinek
2015-02-14 17:29 ` Doug Evans
2015-02-14 18:36 ` Pedro Alves
2015-02-14 22:46 ` Doug Evans
2015-02-09 23:21 ` [PATCH 05/36] Fix redefinition errors in C++ mode Pedro Alves
2015-02-11 10:09 ` Yao Qi
2015-02-11 11:30 ` Pedro Alves
2015-02-11 11:39 ` [PATCH] xcoffread.c: delete 'within_function' definition (Re: [PATCH 05/36] Fix redefinition errors in C++ mode) Pedro Alves
2015-02-09 23:21 ` [PATCH 12/36] proc-service, extern "C" Pedro Alves
2015-02-09 23:21 ` [PATCH 31/36] Split TRY_CATCH into TRY + CATCH Pedro Alves
2015-03-07 15:58 ` Pedro Alves
2015-02-09 23:21 ` [PATCH 15/36] Don't forward declare enum target_hw_bp_type Pedro Alves
2015-02-09 23:21 ` [PATCH 23/36] gdbarch.h: include regcache.h Pedro Alves
2015-02-09 23:21 ` [PATCH 22/36] Remove duplicate const Pedro Alves
2015-02-09 23:22 ` [PATCH 32/36] TRY_CATCH -> TRY+CATCH+END_CATCH everywhere Pedro Alves
2015-02-09 23:22 ` Pedro Alves [this message]
2015-02-09 23:22 ` [PATCH 33/36] TRY_CATCH -> TRY+CATCH+END_CATCH, the manual conversions Pedro Alves
2015-02-09 23:35 ` [PATCH 18/36] Rename struct lzma_stream to avoid clash with system header Pedro Alves
2015-02-09 23:45 ` [PATCH 06/36] record-btrace.c: Remove redefinitions Pedro Alves
2015-02-09 23:46 ` [PATCH 35/36] kill volatile struct gdb_exception Pedro Alves
2015-02-09 23:49 ` [PATCH 00/36] Support building GDB as a C++ program Doug Evans
2015-02-09 23:50 ` [PATCH 34/36] more making TRY/CATCH callers look more like real C++ try/catch blocks Pedro Alves
2015-03-07 15:59 ` Pedro Alves
2015-02-09 23:51 ` [PATCH 07/36] Make array object extern Pedro Alves
2015-02-27 22:47 ` Simon Marchi
2015-02-27 22:58 ` Pedro Alves
2015-02-27 23:04 ` Simon Marchi
2015-02-09 23:51 ` [PATCH 14/36] Do not do arithmetic on enum types Pedro Alves
2015-02-09 23:53 ` [PATCH 01/36] Create libiberty/libiberty.m4, have GDB and GDBserver use it Pedro Alves
2015-02-09 23:35 ` Pedro Alves
2015-02-27 16:23 ` Pedro Alves
2015-02-09 23:54 ` [PATCH 30/36] quit_force: Replace TRY_CATCH wrapper macros Pedro Alves
2015-02-09 23:54 ` [PATCH 10/36] Add extern "C" to declarations of C symbols Pedro Alves
2015-02-11 11:51 ` Pedro Alves
2015-02-10 0:21 ` [PATCH 24/36] breakpoint.h: move enum ‘print_stop_action’ Pedro Alves
2015-02-11 12:28 ` Yao Qi
2015-02-10 15:07 ` [PATCH 00/36] Support building GDB as a C++ program Michael Eager
2015-02-11 17:15 ` Yao Qi
2015-02-12 11:34 ` Pedro Alves
2015-02-17 23:19 ` Patrick Palka
2015-02-18 21:54 ` Yao Qi
2015-02-18 23:49 ` Patrick Palka
2015-02-27 18:19 ` Pedro Alves
2015-02-27 23:52 ` Patrick Palka
2015-02-28 0:09 ` Pedro Alves
2015-03-07 16:01 ` Pedro Alves
2015-03-07 17:58 ` [all pushed] " Pedro Alves
2015-03-16 4:42 ` asmwarrior
2015-03-16 5:05 ` asmwarrior
2015-03-16 5:22 ` asmwarrior
2015-03-16 7:15 ` asmwarrior
2015-03-16 8:20 ` asmwarrior
2015-03-16 11:43 ` [pushed] stub-termcap.c: prototype tputs's parameter's parameter, for C++ mode (Re: [all pushed] Re: [PATCH 00/36] Support building GDB as a C++ program) Pedro Alves
2015-03-16 11:46 ` [pushed] stub termcap, add extern "C" " Pedro Alves
2015-03-16 11:55 ` [all pushed] Re: [PATCH 00/36] Support building GDB as a C++ program Pedro Alves
2015-03-16 11:42 ` Pedro Alves
2015-05-15 8:13 ` asmwarrior
2015-05-15 8:24 ` asmwarrior
2015-05-15 15:09 ` Pedro Alves
2015-05-15 15:04 ` Pedro Alves
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1423524046-20605-37-git-send-email-palves@redhat.com \
--to=palves@redhat.com \
--cc=gdb-patches@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox