Re: [gem5-dev] use of C++ exceptions

[Andreas Hansson]

Google’s C++ guide nicely sums it up:
https://google.github.io/styleguide/cppguide.html#Exceptions

I am not against using exceptions as long as they are only used inside
modules, after careful consideration, and never leaks into the public API
of a module/block.

Andreas

On 10/05/2017, 22:37, "gem5-dev on behalf of nathan binkert"
 wrote:

>This is super historical.  We chose not to use exceptions because we were
>worried that they were "slow".  That's largely just not true anymore.
>IMHO, exceptions definitely have their place.  They're super useful when
>dealing with rare error cases that need to fail a deep call tree.  They
>can
>reduce the number of lines of code by a ton.  There are pitfalls and
>people
>need to know what they're doing (i.e. don't throw in a destructor).
>
>  Nate
>
>On Wed, May 10, 2017 at 9:36 AM, Andreas Sandberg
>
>wrote:
>
>> Having had a quick look at the code, I'd say that exceptions could
>> definitely make sense. I would support limited use of exceptions where
>> it makes sense to make the code flow less entangled. Initially, I would
>> argue that we should keep exceptions local to SimObjects (interfaces
>> like ports should never have to deal with exceptions) and only be used
>> for recoverable errors (e.g., unexpected input that should result in a
>> warning).
>>
>> Are there any technical reason why we aren't using exceptions?
>>
>> //Andreas
>>
>>
>> On 10/05/2017 08:25, Gabe Black wrote:
>>
>>> This would be internal to the gdb code, ie if a read from the socket
>>> fails,
>>> it would detach and throw an exception which would unwind back out of
>>>all
>>> the gdb stuff without having to add ifs all over the place. This bit of
>>> code doesn't really have an external interface, so it wouldn't be
>>>visible
>>> to the caller, assuming a stray exception didn't escape somehow. I
>>>think
>>> it
>>> would be a little nicer that way, but not so much that I'd want to
>>>argue
>>> for it very strongly.
>>>
>>> Gabe
>>>
>>> On Wed, May 10, 2017 at 12:09 AM, Andreas Hansson <
>>> andreas.hans...@arm.com>
>>> wrote:
>>>
>>> Hi Gabe,

 I do not think adding exceptions will make things any less cluttered.
It
 will simply move that complexity to any caller, will it not? I am not
a
 fan of exceptions in general as it mucks with the control flow.

 Andreas

 On 10/05/2017, 07:31, "gem5-dev on behalf of Gabe Black"
  wrote:

 Hi folks. I have a change to make the GDB stub in gem5 a bit less
> fragile:
>
> https://gem5-review.googlesource.com/#/c/3180/
>
> Unfortunately that involved adding a lot of error code checking which
> makes
> things a bit cluttered and ugly. I think it would be a lot nicer to
>use
> exceptions, but I remember those being a no-no. Are they currently
> against
> the rules, or could I use them to make that code a bit nicer?
>
> Gabe
> ___
> gem5-dev mailing list
> gem5-dev@gem5.org
> http://m5sim.org/mailman/listinfo/gem5-dev
>
 IMPORTANT NOTICE: The contents of this email and any attachments are
 confidential and may also be privileged. If you are not the intended
 recipient, please notify the sender immediately and do not disclose
the
 contents to any other person, use it for any purpose, or store or copy
 the
 information in any medium. Thank you.
 ___
 gem5-dev mailing list
 gem5-dev@gem5.org
 http://m5sim.org/mailman/listinfo/gem5-dev

>>> ___
>>> gem5-dev mailing list
>>> gem5-dev@gem5.org
>>> http://m5sim.org/mailman/listinfo/gem5-dev
>>>
>>
>> IMPORTANT NOTICE: The contents of this email and any attachments are
>> confidential and may also be privileged. If you are not the intended
>> recipient, please notify the sender immediately and do not disclose the
>> contents to any other person, use it for any purpose, or store or copy
>>the
>> information in any medium. Thank you.
>> ___
>> gem5-dev mailing list
>> gem5-dev@gem5.org
>> http://m5sim.org/mailman/listinfo/gem5-dev
>>
>___
>gem5-dev mailing list
>gem5-dev@gem5.org
>http://m5sim.org/mailman/listinfo/gem5-dev

IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] use of C++ exceptions

[nathan binkert]

This is super historical.  We chose not to use exceptions because we were
worried that they were "slow".  That's largely just not true anymore.
IMHO, exceptions definitely have their place.  They're super useful when
dealing with rare error cases that need to fail a deep call tree.  They can
reduce the number of lines of code by a ton.  There are pitfalls and people
need to know what they're doing (i.e. don't throw in a destructor).

  Nate

On Wed, May 10, 2017 at 9:36 AM, Andreas Sandberg 
wrote:

> Having had a quick look at the code, I'd say that exceptions could
> definitely make sense. I would support limited use of exceptions where
> it makes sense to make the code flow less entangled. Initially, I would
> argue that we should keep exceptions local to SimObjects (interfaces
> like ports should never have to deal with exceptions) and only be used
> for recoverable errors (e.g., unexpected input that should result in a
> warning).
>
> Are there any technical reason why we aren't using exceptions?
>
> //Andreas
>
>
> On 10/05/2017 08:25, Gabe Black wrote:
>
>> This would be internal to the gdb code, ie if a read from the socket
>> fails,
>> it would detach and throw an exception which would unwind back out of all
>> the gdb stuff without having to add ifs all over the place. This bit of
>> code doesn't really have an external interface, so it wouldn't be visible
>> to the caller, assuming a stray exception didn't escape somehow. I think
>> it
>> would be a little nicer that way, but not so much that I'd want to argue
>> for it very strongly.
>>
>> Gabe
>>
>> On Wed, May 10, 2017 at 12:09 AM, Andreas Hansson <
>> andreas.hans...@arm.com>
>> wrote:
>>
>> Hi Gabe,
>>>
>>> I do not think adding exceptions will make things any less cluttered. It
>>> will simply move that complexity to any caller, will it not? I am not a
>>> fan of exceptions in general as it mucks with the control flow.
>>>
>>> Andreas
>>>
>>> On 10/05/2017, 07:31, "gem5-dev on behalf of Gabe Black"
>>>  wrote:
>>>
>>> Hi folks. I have a change to make the GDB stub in gem5 a bit less
 fragile:

 https://gem5-review.googlesource.com/#/c/3180/

 Unfortunately that involved adding a lot of error code checking which
 makes
 things a bit cluttered and ugly. I think it would be a lot nicer to use
 exceptions, but I remember those being a no-no. Are they currently
 against
 the rules, or could I use them to make that code a bit nicer?

 Gabe
 ___
 gem5-dev mailing list
 gem5-dev@gem5.org
 http://m5sim.org/mailman/listinfo/gem5-dev

>>> IMPORTANT NOTICE: The contents of this email and any attachments are
>>> confidential and may also be privileged. If you are not the intended
>>> recipient, please notify the sender immediately and do not disclose the
>>> contents to any other person, use it for any purpose, or store or copy
>>> the
>>> information in any medium. Thank you.
>>> ___
>>> gem5-dev mailing list
>>> gem5-dev@gem5.org
>>> http://m5sim.org/mailman/listinfo/gem5-dev
>>>
>> ___
>> gem5-dev mailing list
>> gem5-dev@gem5.org
>> http://m5sim.org/mailman/listinfo/gem5-dev
>>
>
> IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
> ___
> gem5-dev mailing list
> gem5-dev@gem5.org
> http://m5sim.org/mailman/listinfo/gem5-dev
>
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] use of C++ exceptions

[Andreas Sandberg]

Having had a quick look at the code, I'd say that exceptions could
definitely make sense. I would support limited use of exceptions where
it makes sense to make the code flow less entangled. Initially, I would
argue that we should keep exceptions local to SimObjects (interfaces
like ports should never have to deal with exceptions) and only be used
for recoverable errors (e.g., unexpected input that should result in a
warning).

Are there any technical reason why we aren't using exceptions?

//Andreas

On 10/05/2017 08:25, Gabe Black wrote:

This would be internal to the gdb code, ie if a read from the socket fails,
it would detach and throw an exception which would unwind back out of all
the gdb stuff without having to add ifs all over the place. This bit of
code doesn't really have an external interface, so it wouldn't be visible
to the caller, assuming a stray exception didn't escape somehow. I think it
would be a little nicer that way, but not so much that I'd want to argue
for it very strongly.

Gabe

On Wed, May 10, 2017 at 12:09 AM, Andreas Hansson 
wrote:


Hi Gabe,

I do not think adding exceptions will make things any less cluttered. It
will simply move that complexity to any caller, will it not? I am not a
fan of exceptions in general as it mucks with the control flow.

Andreas

On 10/05/2017, 07:31, "gem5-dev on behalf of Gabe Black"
 wrote:


Hi folks. I have a change to make the GDB stub in gem5 a bit less fragile:

https://gem5-review.googlesource.com/#/c/3180/

Unfortunately that involved adding a lot of error code checking which
makes
things a bit cluttered and ugly. I think it would be a lot nicer to use
exceptions, but I remember those being a no-no. Are they currently against
the rules, or could I use them to make that code a bit nicer?

Gabe
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

IMPORTANT NOTICE: The contents of this email and any attachments are
confidential and may also be privileged. If you are not the intended
recipient, please notify the sender immediately and do not disclose the
contents to any other person, use it for any purpose, or store or copy the
information in any medium. Thank you.
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev


IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] use of C++ exceptions

[Gabe Black]

This would be internal to the gdb code, ie if a read from the socket fails,
it would detach and throw an exception which would unwind back out of all
the gdb stuff without having to add ifs all over the place. This bit of
code doesn't really have an external interface, so it wouldn't be visible
to the caller, assuming a stray exception didn't escape somehow. I think it
would be a little nicer that way, but not so much that I'd want to argue
for it very strongly.

Gabe

On Wed, May 10, 2017 at 12:09 AM, Andreas Hansson 
wrote:

> Hi Gabe,
>
> I do not think adding exceptions will make things any less cluttered. It
> will simply move that complexity to any caller, will it not? I am not a
> fan of exceptions in general as it mucks with the control flow.
>
> Andreas
>
> On 10/05/2017, 07:31, "gem5-dev on behalf of Gabe Black"
>  wrote:
>
> >Hi folks. I have a change to make the GDB stub in gem5 a bit less fragile:
> >
> >https://gem5-review.googlesource.com/#/c/3180/
> >
> >Unfortunately that involved adding a lot of error code checking which
> >makes
> >things a bit cluttered and ugly. I think it would be a lot nicer to use
> >exceptions, but I remember those being a no-no. Are they currently against
> >the rules, or could I use them to make that code a bit nicer?
> >
> >Gabe
> >___
> >gem5-dev mailing list
> >gem5-dev@gem5.org
> >http://m5sim.org/mailman/listinfo/gem5-dev
>
> IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
> ___
> gem5-dev mailing list
> gem5-dev@gem5.org
> http://m5sim.org/mailman/listinfo/gem5-dev
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Re: [gem5-dev] use of C++ exceptions

[Andreas Hansson]

Hi Gabe,

I do not think adding exceptions will make things any less cluttered. It
will simply move that complexity to any caller, will it not? I am not a
fan of exceptions in general as it mucks with the control flow.

Andreas

On 10/05/2017, 07:31, "gem5-dev on behalf of Gabe Black"
 wrote:

>Hi folks. I have a change to make the GDB stub in gem5 a bit less fragile:
>
>https://gem5-review.googlesource.com/#/c/3180/
>
>Unfortunately that involved adding a lot of error code checking which
>makes
>things a bit cluttered and ugly. I think it would be a lot nicer to use
>exceptions, but I remember those being a no-no. Are they currently against
>the rules, or could I use them to make that code a bit nicer?
>
>Gabe
>___
>gem5-dev mailing list
>gem5-dev@gem5.org
>http://m5sim.org/mailman/listinfo/gem5-dev

IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

[gem5-dev] use of C++ exceptions

[Gabe Black]

Hi folks. I have a change to make the GDB stub in gem5 a bit less fragile:

https://gem5-review.googlesource.com/#/c/3180/

Unfortunately that involved adding a lot of error code checking which makes
things a bit cluttered and ugly. I think it would be a lot nicer to use
exceptions, but I remember those being a no-no. Are they currently against
the rules, or could I use them to make that code a bit nicer?

Gabe
___
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev