Mirror of the gdb mailing list
 help / color / mirror / Atom feed
* Question Regarding Int Cast Error in GDB Python
@ 2025-08-22 18:21 Vacha Bhavsar via Gdb
  2025-08-25  8:35 ` Luis via Gdb
  2025-08-29 17:46 ` Tom Tromey
  0 siblings, 2 replies; 8+ messages in thread
From: Vacha Bhavsar via Gdb @ 2025-08-22 18:21 UTC (permalink / raw)
  To: gdb

Hi,

I am developing a patch for QEMU which would allow QEMU to expose
SME registers, such as the ZA storage, via gdbstub to a remote
gdb client session. I am working to write a simple test case for
this change which simply sets the value of the za register, then
reads it and ensures what is read is equivalent to what was set.

The full test case can be found here:

https://lists.gnu.org/archive/html/qemu-devel/2025-08/msg01815.html

However, it seems there is an issue in the int cast in the following
code:

v = gdb.parse_and_eval(reg)
report(str(v.type) == "uint8_t", "size of %s" % (reg))
report(int(v) == 0x1, "%s is 0x%x" % (reg, 0x1))

This results in the following error:

Python Exception <class 'gdb.error'>: That operation is not available
on integers of more than 8 bytes.
Error occurred in Python: That operation is not available on integers
of more than 8 bytes.

After running various tests with different host architectures, gdb
versions, python version and int casting included and excluded, we
are seeing that this issue only seems to arise when the int cast is
included and the gdb used is GNU gdb (Ubuntu 15.0.50.20240403-0ubuntu1)
15.0.50.20240403-git (regardless of python version and host architecture).

The test results are presented below. I was hoping to get some guidance
regarding this error and what may be the cause if it indeed stems from
this specific version of gdb?

Thanks,
Vacha


 gdb     gdb target           python   host          int cast    status
version support              version  architecture
16.3 --enable-targets=all 3.11.13   x86        yes        pass
16.3 --enable-targets=all 3.11.13   x86         no        pass
16.3 --enable-targets=all 3.10.18   x86        yes        pass
16.3 --enable-targets=all 3.10.18   x86         no        pass
16.3 --enable-targets=all 3.8.10    x86        yes        pass
16.3 --enable-targets=all 3.8.10    x86          no        pass

16.3   aarch64             3.11.0rc1 aarch64      yes        pass
16.3   aarch64             3.11.0rc1 aarch64       no        pass
16.3   aarch64             3.10.12   aarch64      yes        pass
16.3   aarch64             3.10.12   aarch64       no        pass

15.0   multiarch           3.10.12   aarch64      yes        fail
15.0   multiarch           3.10.12   aarch64      no        pass
15.0   multiarch          3.11.0rc1 aarch64      yes        fail
15.0   multiarch          3.11.0rc1 aarch64      no        pass

15.0   multiarch          3.8.10      x86            yes        fail
15.0   multiarch          3.8.10      x86             no        pass
15.0   multiarch        3.11.13      x86            yes        fail
15.0   multiarch        3.11.13      x86             no        pass
15.0   multiarch        3.10.18      x86            yes        fail
15.0   multiarch        3.10.18     x86             no        pass

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-22 18:21 Question Regarding Int Cast Error in GDB Python Vacha Bhavsar via Gdb
@ 2025-08-25  8:35 ` Luis via Gdb
  2025-08-25 13:23   ` Vacha Bhavsar via Gdb
  2025-08-29 17:46 ` Tom Tromey
  1 sibling, 1 reply; 8+ messages in thread
From: Luis via Gdb @ 2025-08-25  8:35 UTC (permalink / raw)
  To: Vacha Bhavsar, gdb

On 8/22/25 19:21, Vacha Bhavsar via Gdb wrote:
> Hi,
> 
> I am developing a patch for QEMU which would allow QEMU to expose
> SME registers, such as the ZA storage, via gdbstub to a remote
> gdb client session. I am working to write a simple test case for
> this change which simply sets the value of the za register, then
> reads it and ensures what is read is equivalent to what was set.
> 
> The full test case can be found here:
> 
> https://lists.gnu.org/archive/html/qemu-devel/2025-08/msg01815.html
> 
> However, it seems there is an issue in the int cast in the following
> code:
> 
> v = gdb.parse_and_eval(reg)
> report(str(v.type) == "uint8_t", "size of %s" % (reg))
> report(int(v) == 0x1, "%s is 0x%x" % (reg, 0x1))

If reg is ZA, I´d expect gdb to have difficulties casting the whole 
array to an int, hence the message about not managing to operate on 
integers larger than 8 bytes.

How are you testing ZA? Byte by byte?

> 
> This results in the following error:
> 
> Python Exception <class 'gdb.error'>: That operation is not available
> on integers of more than 8 bytes.
> Error occurred in Python: That operation is not available on integers
> of more than 8 bytes.
> 
> After running various tests with different host architectures, gdb
> versions, python version and int casting included and excluded, we
> are seeing that this issue only seems to arise when the int cast is
> included and the gdb used is GNU gdb (Ubuntu 15.0.50.20240403-0ubuntu1)
> 15.0.50.20240403-git (regardless of python version and host architecture).
> 
> The test results are presented below. I was hoping to get some guidance
> regarding this error and what may be the cause if it indeed stems from
> this specific version of gdb?
> 
> Thanks,
> Vacha
> 
> 
>   gdb     gdb target           python   host          int cast    status
> version support              version  architecture
> 16.3 --enable-targets=all 3.11.13   x86        yes        pass
> 16.3 --enable-targets=all 3.11.13   x86         no        pass
> 16.3 --enable-targets=all 3.10.18   x86        yes        pass
> 16.3 --enable-targets=all 3.10.18   x86         no        pass
> 16.3 --enable-targets=all 3.8.10    x86        yes        pass
> 16.3 --enable-targets=all 3.8.10    x86          no        pass
> 
> 16.3   aarch64             3.11.0rc1 aarch64      yes        pass
> 16.3   aarch64             3.11.0rc1 aarch64       no        pass
> 16.3   aarch64             3.10.12   aarch64      yes        pass
> 16.3   aarch64             3.10.12   aarch64       no        pass
> 
> 15.0   multiarch           3.10.12   aarch64      yes        fail
> 15.0   multiarch           3.10.12   aarch64      no        pass
> 15.0   multiarch          3.11.0rc1 aarch64      yes        fail
> 15.0   multiarch          3.11.0rc1 aarch64      no        pass
> 
> 15.0   multiarch          3.8.10      x86            yes        fail
> 15.0   multiarch          3.8.10      x86             no        pass
> 15.0   multiarch        3.11.13      x86            yes        fail
> 15.0   multiarch        3.11.13      x86             no        pass
> 15.0   multiarch        3.10.18      x86            yes        fail
> 15.0   multiarch        3.10.18     x86             no        pass


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-25  8:35 ` Luis via Gdb
@ 2025-08-25 13:23   ` Vacha Bhavsar via Gdb
  2025-08-25 13:29     ` Vacha Bhavsar via Gdb
  0 siblings, 1 reply; 8+ messages in thread
From: Vacha Bhavsar via Gdb @ 2025-08-25 13:23 UTC (permalink / raw)
  To: Luis; +Cc: gdb

Hi,

We are testing the ZA by tiles and tile slices. I've included a larger
snippet
of the code below:

for i in range(0, 4):
        for j in range(0, 4):
            for k in range(0, 4):
                cmd = "set $za%dhq%d[%d] = 0x%x" % (i, j, k, MAGIC)
                gdb.execute(cmd)
                report(True, "%s" % cmd)
        for j in range(0, 4):
            for k in range(0, 4):
                reg = "$za%dhq%d[%d]" % (i, j, k)
                v = gdb.parse_and_eval(reg)
                report(str(v.type) == "uint128_t",
                    "size of %s" % (reg))
                report(int(v) == MAGIC, "%s is 0x%x" % (reg, MAGIC))

So in this case we see the error when we try to test za0hq0[0] when
v is cast to an int.

On Mon, Aug 25, 2025 at 4:35 AM Luis <luis.machado.foss@gmail.com> wrote:

> On 8/22/25 19:21, Vacha Bhavsar via Gdb wrote:
> > Hi,
> >
> > I am developing a patch for QEMU which would allow QEMU to expose
> > SME registers, such as the ZA storage, via gdbstub to a remote
> > gdb client session. I am working to write a simple test case for
> > this change which simply sets the value of the za register, then
> > reads it and ensures what is read is equivalent to what was set.
> >
> > The full test case can be found here:
> >
> > https://lists.gnu.org/archive/html/qemu-devel/2025-08/msg01815.html
> >
> > However, it seems there is an issue in the int cast in the following
> > code:
> >
> > v = gdb.parse_and_eval(reg)
> > report(str(v.type) == "uint8_t", "size of %s" % (reg))
> > report(int(v) == 0x1, "%s is 0x%x" % (reg, 0x1))
>
> If reg is ZA, I´d expect gdb to have difficulties casting the whole
> array to an int, hence the message about not managing to operate on
> integers larger than 8 bytes.
>
> How are you testing ZA? Byte by byte?
>
> >
> > This results in the following error:
> >
> > Python Exception <class 'gdb.error'>: That operation is not available
> > on integers of more than 8 bytes.
> > Error occurred in Python: That operation is not available on integers
> > of more than 8 bytes.
> >
> > After running various tests with different host architectures, gdb
> > versions, python version and int casting included and excluded, we
> > are seeing that this issue only seems to arise when the int cast is
> > included and the gdb used is GNU gdb (Ubuntu 15.0.50.20240403-0ubuntu1)
> > 15.0.50.20240403-git (regardless of python version and host
> architecture).
> >
> > The test results are presented below. I was hoping to get some guidance
> > regarding this error and what may be the cause if it indeed stems from
> > this specific version of gdb?
> >
> > Thanks,
> > Vacha
> >
> >
> >   gdb     gdb target           python   host          int cast    status
> > version support              version  architecture
> > 16.3 --enable-targets=all 3.11.13   x86        yes        pass
> > 16.3 --enable-targets=all 3.11.13   x86         no        pass
> > 16.3 --enable-targets=all 3.10.18   x86        yes        pass
> > 16.3 --enable-targets=all 3.10.18   x86         no        pass
> > 16.3 --enable-targets=all 3.8.10    x86        yes        pass
> > 16.3 --enable-targets=all 3.8.10    x86          no        pass
> >
> > 16.3   aarch64             3.11.0rc1 aarch64      yes        pass
> > 16.3   aarch64             3.11.0rc1 aarch64       no        pass
> > 16.3   aarch64             3.10.12   aarch64      yes        pass
> > 16.3   aarch64             3.10.12   aarch64       no        pass
> >
> > 15.0   multiarch           3.10.12   aarch64      yes        fail
> > 15.0   multiarch           3.10.12   aarch64      no        pass
> > 15.0   multiarch          3.11.0rc1 aarch64      yes        fail
> > 15.0   multiarch          3.11.0rc1 aarch64      no        pass
> >
> > 15.0   multiarch          3.8.10      x86            yes        fail
> > 15.0   multiarch          3.8.10      x86             no        pass
> > 15.0   multiarch        3.11.13      x86            yes        fail
> > 15.0   multiarch        3.11.13      x86             no        pass
> > 15.0   multiarch        3.10.18      x86            yes        fail
> > 15.0   multiarch        3.10.18     x86             no        pass
>
>

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-25 13:23   ` Vacha Bhavsar via Gdb
@ 2025-08-25 13:29     ` Vacha Bhavsar via Gdb
  2025-08-26 20:49       ` Luis via Gdb
  0 siblings, 1 reply; 8+ messages in thread
From: Vacha Bhavsar via Gdb @ 2025-08-25 13:29 UTC (permalink / raw)
  To: Luis; +Cc: gdb

Hi,

Apologies, it seems I pasted the wrong snippet in the original email.
The code in my reply just prior to this is the correct code. The snippet
that
should have been included in the first email, which causes the error, is
the one below:

v = gdb.parse_and_eval(reg)
report(str(v.type) == "uint128_t", "size of %s" % (reg))
report(int(v) == MAGIC, "%s is 0x%x" % (reg, MAGIC))

Sorry for the confusion!


On Mon, Aug 25, 2025 at 9:23 AM Vacha Bhavsar <
vacha.bhavsar@oss.qualcomm.com> wrote:

> Hi,
>
> We are testing the ZA by tiles and tile slices. I've included a larger
> snippet
> of the code below:
>
> for i in range(0, 4):
>         for j in range(0, 4):
>             for k in range(0, 4):
>                 cmd = "set $za%dhq%d[%d] = 0x%x" % (i, j, k, MAGIC)
>                 gdb.execute(cmd)
>                 report(True, "%s" % cmd)
>         for j in range(0, 4):
>             for k in range(0, 4):
>                 reg = "$za%dhq%d[%d]" % (i, j, k)
>                 v = gdb.parse_and_eval(reg)
>                 report(str(v.type) == "uint128_t",
>                     "size of %s" % (reg))
>                 report(int(v) == MAGIC, "%s is 0x%x" % (reg, MAGIC))
>
> So in this case we see the error when we try to test za0hq0[0] when
> v is cast to an int.
>
> On Mon, Aug 25, 2025 at 4:35 AM Luis <luis.machado.foss@gmail.com> wrote:
>
>> On 8/22/25 19:21, Vacha Bhavsar via Gdb wrote:
>> > Hi,
>> >
>> > I am developing a patch for QEMU which would allow QEMU to expose
>> > SME registers, such as the ZA storage, via gdbstub to a remote
>> > gdb client session. I am working to write a simple test case for
>> > this change which simply sets the value of the za register, then
>> > reads it and ensures what is read is equivalent to what was set.
>> >
>> > The full test case can be found here:
>> >
>> > https://lists.gnu.org/archive/html/qemu-devel/2025-08/msg01815.html
>> >
>> > However, it seems there is an issue in the int cast in the following
>> > code:
>> >
>> > v = gdb.parse_and_eval(reg)
>> > report(str(v.type) == "uint8_t", "size of %s" % (reg))
>> > report(int(v) == 0x1, "%s is 0x%x" % (reg, 0x1))
>>
>> If reg is ZA, I´d expect gdb to have difficulties casting the whole
>> array to an int, hence the message about not managing to operate on
>> integers larger than 8 bytes.
>>
>> How are you testing ZA? Byte by byte?
>>
>> >
>> > This results in the following error:
>> >
>> > Python Exception <class 'gdb.error'>: That operation is not available
>> > on integers of more than 8 bytes.
>> > Error occurred in Python: That operation is not available on integers
>> > of more than 8 bytes.
>> >
>> > After running various tests with different host architectures, gdb
>> > versions, python version and int casting included and excluded, we
>> > are seeing that this issue only seems to arise when the int cast is
>> > included and the gdb used is GNU gdb (Ubuntu 15.0.50.20240403-0ubuntu1)
>> > 15.0.50.20240403-git (regardless of python version and host
>> architecture).
>> >
>> > The test results are presented below. I was hoping to get some guidance
>> > regarding this error and what may be the cause if it indeed stems from
>> > this specific version of gdb?
>> >
>> > Thanks,
>> > Vacha
>> >
>> >
>> >   gdb     gdb target           python   host          int cast    status
>> > version support              version  architecture
>> > 16.3 --enable-targets=all 3.11.13   x86        yes        pass
>> > 16.3 --enable-targets=all 3.11.13   x86         no        pass
>> > 16.3 --enable-targets=all 3.10.18   x86        yes        pass
>> > 16.3 --enable-targets=all 3.10.18   x86         no        pass
>> > 16.3 --enable-targets=all 3.8.10    x86        yes        pass
>> > 16.3 --enable-targets=all 3.8.10    x86          no        pass
>> >
>> > 16.3   aarch64             3.11.0rc1 aarch64      yes        pass
>> > 16.3   aarch64             3.11.0rc1 aarch64       no        pass
>> > 16.3   aarch64             3.10.12   aarch64      yes        pass
>> > 16.3   aarch64             3.10.12   aarch64       no        pass
>> >
>> > 15.0   multiarch           3.10.12   aarch64      yes        fail
>> > 15.0   multiarch           3.10.12   aarch64      no        pass
>> > 15.0   multiarch          3.11.0rc1 aarch64      yes        fail
>> > 15.0   multiarch          3.11.0rc1 aarch64      no        pass
>> >
>> > 15.0   multiarch          3.8.10      x86            yes        fail
>> > 15.0   multiarch          3.8.10      x86             no        pass
>> > 15.0   multiarch        3.11.13      x86            yes        fail
>> > 15.0   multiarch        3.11.13      x86             no        pass
>> > 15.0   multiarch        3.10.18      x86            yes        fail
>> > 15.0   multiarch        3.10.18     x86             no        pass
>>
>>

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-25 13:29     ` Vacha Bhavsar via Gdb
@ 2025-08-26 20:49       ` Luis via Gdb
  2025-08-27 19:16         ` Vacha Bhavsar via Gdb
  0 siblings, 1 reply; 8+ messages in thread
From: Luis via Gdb @ 2025-08-26 20:49 UTC (permalink / raw)
  To: Vacha Bhavsar; +Cc: gdb

Hi,

On 8/25/25 14:29, Vacha Bhavsar wrote:
> Hi,
> 
> Apologies, it seems I pasted the wrong snippet in the original email.
> The code in my reply just prior to this is the correct code. The snippet 
> that
> should have been included in the first email, which causes the error, is
> the one below:
> 
> v = gdb.parse_and_eval(reg)
> report(str(v.type) == "uint128_t", "size of %s" % (reg))

Right, so gdb doesn't like casting integers to/from 8+ bytes. That's a 
limitation currently. But why are you considering this particular cast? 
To optimize the checks instead of using the byte vectors?

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-26 20:49       ` Luis via Gdb
@ 2025-08-27 19:16         ` Vacha Bhavsar via Gdb
  0 siblings, 0 replies; 8+ messages in thread
From: Vacha Bhavsar via Gdb @ 2025-08-27 19:16 UTC (permalink / raw)
  To: Luis; +Cc: gdb

Hi,

We chose this particular cast just to stay aligned with the
existing method of testing in QEMU which also uses this cast
but for 64bit integers.

qemu/tests/tcg/aarch64/gdbstub/test-sve.py at
b86307ecef9222c335ebd0ed4da2b243e86f779e · qemu/qemu
<https://github.com/qemu/qemu/blob/b86307ecef9222c335ebd0ed4da2b243e86f779e/tests/tcg/aarch64/gdbstub/test-sve.py>
https://lists.gnu.org/archive/html/qemu-devel/2025-08/msg01815.html


However, since we have seen that this issue only occurs with the
specific indiciated version of gdb, and does not fail for example with
gdb16.3, we have implemented a workaround to avoid the
casting in the case of gdb-multiarch 15.0.50.20240403-0ubuntu1.

Thank you for confirming this is a known issue!


Thanks,
Vacha

On Tue, Aug 26, 2025 at 4:49 PM Luis <luis.machado.foss@gmail.com> wrote:

> Hi,
>
> On 8/25/25 14:29, Vacha Bhavsar wrote:
> > Hi,
> >
> > Apologies, it seems I pasted the wrong snippet in the original email.
> > The code in my reply just prior to this is the correct code. The snippet
> > that
> > should have been included in the first email, which causes the error, is
> > the one below:
> >
> > v = gdb.parse_and_eval(reg)
> > report(str(v.type) == "uint128_t", "size of %s" % (reg))
>
> Right, so gdb doesn't like casting integers to/from 8+ bytes. That's a
> limitation currently. But why are you considering this particular cast?
> To optimize the checks instead of using the byte vectors?
>

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-22 18:21 Question Regarding Int Cast Error in GDB Python Vacha Bhavsar via Gdb
  2025-08-25  8:35 ` Luis via Gdb
@ 2025-08-29 17:46 ` Tom Tromey
  2025-09-03 18:59   ` Vacha Bhavsar via Gdb
  1 sibling, 1 reply; 8+ messages in thread
From: Tom Tromey @ 2025-08-29 17:46 UTC (permalink / raw)
  To: Vacha Bhavsar via Gdb; +Cc: Vacha Bhavsar

>>>>> ">" == Vacha Bhavsar via Gdb <gdb@sourceware.org> writes:

>> v = gdb.parse_and_eval(reg)
>> report(str(v.type) == "uint8_t", "size of %s" % (reg))
>> report(int(v) == 0x1, "%s is 0x%x" % (reg, 0x1))

>> This results in the following error:

>> Python Exception <class 'gdb.error'>: That operation is not available
>> on integers of more than 8 bytes.

I think this is just a bug.  Would you mind filing it in bugzilla?

python/py-value.c:valpy_long uses value_as_long.  However if it used
value_as_mpz instead, it could handle types wider than 64 bits.

Tom

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: Question Regarding Int Cast Error in GDB Python
  2025-08-29 17:46 ` Tom Tromey
@ 2025-09-03 18:59   ` Vacha Bhavsar via Gdb
  0 siblings, 0 replies; 8+ messages in thread
From: Vacha Bhavsar via Gdb @ 2025-09-03 18:59 UTC (permalink / raw)
  To: Tom Tromey; +Cc: Vacha Bhavsar via Gdb

I have filed this in bugzilla.

Thanks,
Vacha

On Fri, Aug 29, 2025 at 1:46 PM Tom Tromey <tom@tromey.com> wrote:

> >>>>> ">" == Vacha Bhavsar via Gdb <gdb@sourceware.org> writes:
>
> >> v = gdb.parse_and_eval(reg)
> >> report(str(v.type) == "uint8_t", "size of %s" % (reg))
> >> report(int(v) == 0x1, "%s is 0x%x" % (reg, 0x1))
>
> >> This results in the following error:
>
> >> Python Exception <class 'gdb.error'>: That operation is not available
> >> on integers of more than 8 bytes.
>
> I think this is just a bug.  Would you mind filing it in bugzilla?
>
> python/py-value.c:valpy_long uses value_as_long.  However if it used
> value_as_mpz instead, it could handle types wider than 64 bits.
>
> Tom
>

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2025-09-03 19:00 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-08-22 18:21 Question Regarding Int Cast Error in GDB Python Vacha Bhavsar via Gdb
2025-08-25  8:35 ` Luis via Gdb
2025-08-25 13:23   ` Vacha Bhavsar via Gdb
2025-08-25 13:29     ` Vacha Bhavsar via Gdb
2025-08-26 20:49       ` Luis via Gdb
2025-08-27 19:16         ` Vacha Bhavsar via Gdb
2025-08-29 17:46 ` Tom Tromey
2025-09-03 18:59   ` Vacha Bhavsar via Gdb

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox