[PATCH PR gdb/18071] TLS variables can't be resolved on aarch64-linux-gnu

Wei-min Pan weimin.pan@oracle.com
Thu Mar 22 18:18:00 GMT 2018


Hi Simon,

On 3/21/2018 9:22 PM, Simon Marchi wrote:
> On 2018-03-19 15:36, Weimin Pan wrote:
>> Yes, printing of a TLS fails on all platforms, not just on aarch64.
>> How about changing the title to:
>>
>>    [PATCH PR gdb/18071] aarch64: "info" command can't resolve TLS 
>> variables
>
> Maybe "info address" instead of "info"?

OK.

>
>>> Is this code equivalent to calling lookup_minimal_symbol (name, 
>>> NULL, NULL) ?  If
>>> so, there's already lookup_bound_minimal_symbol that does the same, 
>>> so maybe we
>>> can just drop lookup_minimal_symbol_and_objfile and use 
>>> lookup_bound_minimal_symbol.
>>
>> Yes, it turns out lookup_minimal_symbol_and_objfile(name) is 
>> equivalent to
>> calling lookup_minimal_symbol (name, NULL, NULL). We can replace the 
>> call in
>> info_address_command() with either lookup_minimal_symbol (name, NULL, 
>> NULL)
>> or lookup_bound_minimal_symbol(name). Calling either one looks like a
>> better fix.
>>
>>>
>>> We could also just have lookup_minimal_symbol with parameters that 
>>> default to nullptr.
>>> It is not clear at all to have lookup_bound_minimal_symbol and 
>>> lookup_minimal_symbol
>>> that both return a bound_minimal_symbol, that's quite misleading.
>>
>> I don't know the rational behind having these two functions which get 
>> called in
>> quite a few places. Yes, having default arguments for 
>> lookup_minimal_symbol()
>> will be another option.
>>
>> Thanks very much for your comments.
>
> Thanks for your patience :).  This refactoring can be done in a 
> separate patch, to keep this one focused on fixing the bug.

Thank you very much for getting this process moving :) Would you like me 
to submit a revised patch which calls
lookup_bound_minimal_symbol instead of lookup_minimal_symbol_and_objfile?

Weimin

>
> Simon



More information about the Gdb-patches mailing list