The refactoring of handle_set_reg missed the fact we previously had
responded with an empty packet when we were not using XML based
protocols. This broke the fallback behaviour for architectures that
don't have registers defined in QEMU's gdb-xml directory.

Revert to the previous behaviour and clean up the commentary for what
is going on.

Fixes: 62b3320bddd
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Cc: Jon Doron <arilou@gmail.com>
Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
---
 gdbstub.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/gdbstub.c b/gdbstub.c
index ea3349d1aa..b6df7ee25a 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -1669,12 +1669,23 @@ static void handle_remove_bp(GdbCmdContext *gdb_ctx, void *user_ctx)
     put_packet(gdb_ctx->s, "E22");
 }
 
+/*
+ * handle_set/get_reg
+ *
+ * Older gdb are really dumb, and don't use 'G/g' if 'P/p' is available.
+ * This works, but can be very slow. Anything new enough to understand
+ * XML also knows how to use this properly. However to use this we
+ * need to define a local XML file as well as be talking to a
+ * reasonably modern gdb. Responding with an empty packet will cause
+ * the remote gdb to fallback to older methods.
+ */
+
 static void handle_set_reg(GdbCmdContext *gdb_ctx, void *user_ctx)
 {
     int reg_size;
 
     if (!gdb_has_xml) {
-        put_packet(gdb_ctx->s, "E00");
+        put_packet(gdb_ctx->s, "");
         return;
     }
 
@@ -1694,11 +1705,6 @@ static void handle_get_reg(GdbCmdContext *gdb_ctx, void *user_ctx)
 {
     int reg_size;
 
-    /*
-     * Older gdb are really dumb, and don't use 'g' if 'p' is avaialable.
-     * This works, but can be very slow.  Anything new enough to
-     * understand XML also knows how to use this properly.
-     */
     if (!gdb_has_xml) {
         put_packet(gdb_ctx->s, "");
         return;
-- 
2.20.1

On 05/07/2019 14:29, Alex Bennée wrote:

> The refactoring of handle_set_reg missed the fact we previously had
> responded with an empty packet when we were not using XML based
> protocols. This broke the fallback behaviour for architectures that
> don't have registers defined in QEMU's gdb-xml directory.
> 
> Revert to the previous behaviour and clean up the commentary for what
> is going on.
> 
> Fixes: 62b3320bddd
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> Cc: Jon Doron <arilou@gmail.com>
> Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
> ---
>  gdbstub.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
> 
> diff --git a/gdbstub.c b/gdbstub.c
> index ea3349d1aa..b6df7ee25a 100644
> --- a/gdbstub.c
> +++ b/gdbstub.c
> @@ -1669,12 +1669,23 @@ static void handle_remove_bp(GdbCmdContext *gdb_ctx, void *user_ctx)
>      put_packet(gdb_ctx->s, "E22");
>  }
>  
> +/*
> + * handle_set/get_reg
> + *
> + * Older gdb are really dumb, and don't use 'G/g' if 'P/p' is available.
> + * This works, but can be very slow. Anything new enough to understand
> + * XML also knows how to use this properly. However to use this we
> + * need to define a local XML file as well as be talking to a
> + * reasonably modern gdb. Responding with an empty packet will cause
> + * the remote gdb to fallback to older methods.
> + */
> +
>  static void handle_set_reg(GdbCmdContext *gdb_ctx, void *user_ctx)
>  {
>      int reg_size;
>  
>      if (!gdb_has_xml) {
> -        put_packet(gdb_ctx->s, "E00");
> +        put_packet(gdb_ctx->s, "");
>          return;
>      }
>  
> @@ -1694,11 +1705,6 @@ static void handle_get_reg(GdbCmdContext *gdb_ctx, void *user_ctx)
>  {
>      int reg_size;
>  
> -    /*
> -     * Older gdb are really dumb, and don't use 'g' if 'p' is avaialable.
> -     * This works, but can be very slow.  Anything new enough to
> -     * understand XML also knows how to use this properly.
> -     */
>      if (!gdb_has_xml) {
>          put_packet(gdb_ctx->s, "");
>          return;
> 

Works for me - thanks for the quick response!

Tested-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>


ATB,

Mark.

On 7/5/19 3:29 PM, Alex Bennée wrote:
> The refactoring of handle_set_reg missed the fact we previously had
> responded with an empty packet when we were not using XML based
> protocols. This broke the fallback behaviour for architectures that
> don't have registers defined in QEMU's gdb-xml directory.
> 
> Revert to the previous behaviour and clean up the commentary for what
> is going on.
> 
> Fixes: 62b3320bddd
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> Cc: Jon Doron <arilou@gmail.com>
> Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
> ---
>  gdbstub.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
> 
> diff --git a/gdbstub.c b/gdbstub.c
> index ea3349d1aa..b6df7ee25a 100644
> --- a/gdbstub.c
> +++ b/gdbstub.c
> @@ -1669,12 +1669,23 @@ static void handle_remove_bp(GdbCmdContext *gdb_ctx, void *user_ctx)
>      put_packet(gdb_ctx->s, "E22");
>  }
>  
> +/*
> + * handle_set/get_reg
> + *
> + * Older gdb are really dumb, and don't use 'G/g' if 'P/p' is available.
> + * This works, but can be very slow. Anything new enough to understand
> + * XML also knows how to use this properly. However to use this we
> + * need to define a local XML file as well as be talking to a
> + * reasonably modern gdb. Responding with an empty packet will cause
> + * the remote gdb to fallback to older methods.
> + */
> +
>  static void handle_set_reg(GdbCmdContext *gdb_ctx, void *user_ctx)
>  {
>      int reg_size;
>  
>      if (!gdb_has_xml) {
> -        put_packet(gdb_ctx->s, "E00");
> +        put_packet(gdb_ctx->s, "");
>          return;
>      }
>  
> @@ -1694,11 +1705,6 @@ static void handle_get_reg(GdbCmdContext *gdb_ctx, void *user_ctx)
>  {
>      int reg_size;
>  
> -    /*
> -     * Older gdb are really dumb, and don't use 'g' if 'p' is avaialable.
> -     * This works, but can be very slow.  Anything new enough to
> -     * understand XML also knows how to use this properly.
> -     */
>      if (!gdb_has_xml) {
>          put_packet(gdb_ctx->s, "");
>          return;
> 

Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>