Re: [Libunwind-devel] use_prev_instr breaking unwind - how to stop it?

[Doug Moore]

It’s not clear to me how to push this to you via GitHub.  So, instead, I’ve 
attached a patch to add unw_local_init_signal to the API, to solve my problem.  
Please consider it.

Doug

add_local_init_signal.patch
Description: Binary data

> On Mar 30, 2017, at 12:00 PM, Doug Moore <address@hidden> wrote:
> 
> Documentation has to change anyway, since the unw_init_local docs are 
> currently broken.
> 
> Decrementing the ip for my case (using unw_set_reg) gets me through a few 
> unw_step calls successfully, then crashes the program.
> 
> But I can't win this fight, so I'll quit.
> 
> Maybe I can get a new cursor init function added to the API.  Would the 
> powers that be accept a change that added unw_init_local_signal, which is 
> just like unw_init_local except that it doesn't set use_prev_instr?  If not 
> that name, the same thing by some other name?  I can post that change pretty 
> quickly, if only someone will consider it.
> 
> Doug
> 
> 
> Quoting Dave Watson <address@hidden>:
> 
>> On 03/29/17 04:51 PM, Doug Moore wrote:
>>> Can’t we fix this whole problem by having getcontext subtract from the pc 
>>> to start with, so that nobody ever has to decrement the pc on the first 
>>> call to unw_step?
>> 
>> That would assume that the ucontext is usually from a signal frame
>> (third argument from a sigaction handler), which is probably pretty
>> common.  It would break anyone that grabbed it outside a handler using
>> getcontext() instead of unw_getcontext() though.
>> 
>> The current documentation says:
>> 
>> ```
>> On IA-64, unw_context_t has a layout that is compatible with that of
>> ucontext_t and such structures can be initialized with getcontext()
>> instead of unw_getcontext()
>> ```
>> 
>> so it might be a somewhat interface breaking change.
>> 
>> Conversely, for your use case you might just be able to bump the ip by
>> one in the ucontext before the first unw_step()?
>> 
>> 
>> !DSPAM:10223,58dd225839344861828991!
> 
>