[lldb-dev] UnicodeDecodeError for serialize SBValue description

Enrico Granata via lldb-dev lldb-dev at lists.llvm.org
Thu Apr 7 10:33:49 PDT 2016


> On Apr 6, 2016, at 7:31 PM, Jeffrey Tan <jeffrey.fudan at gmail.com> wrote:
> 
> Thanks Enrico. This is very detailed! I will take a look. 
> Btw: originally, I was hoping that data formatter can be added without changing the source code. Like giving a xml/json format file telling lldb the memory layout/structure of the data structure, lldb can parse the xml/json and deduce the formatting. This is approach used by data visualizer in VS debugger: https://msdn.microsoft.com/en-us/library/jj620914.aspx <https://msdn.microsoft.com/en-us/library/jj620914.aspx>
> This will make adding data formatter more extensible/flexible. Any reason we did not take this approach? 
> 

The way I understand the Natvis system, it allows one to provide a bunch of expressions that describe how the debugger would go about retrieving the interesting data bits
This has the bonus of being really easy, since you’re writing code in the same language/context of the types you’re formatting
On the other hand it has a few drawbacks, in terms of performance as well as safety (imagine trying to run code on an object when said object is in an incoherent state)
The LLDB approach, on the other hand, is that you should try to not run code when providing these data formatters. In order to do that, we vend an API that can do things such as retrieve child values, read memory, cast values, …, all without code execution
Once you have this kind of API that is not expressed in your source language, you might just as well describe it in a scripting language. Hence were born the Python data formatters.
In order for us to gain even more performance for native system types that we know we’re gonna run into all the time, we then switched a bunch of the “mission critical” formatters from Python to C++
The Python extension points are still available, as Jim pointed out, and you are more than welcome to use those instead of modifying the debugger core

> Jeffrey
> 
> On Wed, Apr 6, 2016 at 11:49 AM, Enrico Granata <egranata at apple.com <mailto:egranata at apple.com>> wrote:
> 
>> On Apr 5, 2016, at 2:42 PM, Jeffrey Tan <jeffrey.fudan at gmail.com <mailto:jeffrey.fudan at gmail.com>> wrote:
>> 
>> Hi Enrico,
>> 
>> Any suggestion/example how to add a data formatter for our own STL string? From the output below I can see we are using our own "fbstring_core" which I assume I need to write a type summary for this type:
>> 
>> frame variable corpus -T
>> (const string &const) corpus = error: summary string parsing error: {
>>   (std::fbstring_core<char>) store_ = {
>>     (std::fbstring_core<char>::(anonymous union))  = {
>>       (char [24]) small_ = "www"
>>       (std::fbstring_core<char>::MediumLarge) ml_ = {
>>         (char *) data_ = 0x0000000000777777 "H\x89U\xa8H\x89M\xa0L\x89E\x98H\x8bE\xa8H\x89��_U��D\x88e�H\x8bE\xa0H\x89��]U��H\x89�H\x8dE�H\x89�H\x89��� ��L\x8dm�H\x8bE\x98H\x89��IU��\x88]�L\x8be\xb0L\x89��
>>         (std::size_t) size_ = 0
>>         (std::size_t) capacity_ = 1441151880758558720
>>       }
>>     }
>>   }
>> }
>> 
> 
> Admittedly, this is going to be a little vague since I haven’t really seen your code and I am only working off of one sample
> 
> There’s going to be two parts to getting this to work:
> 
> Part 1 - Formatting fbstring_core
> 
> At a glance, an fbstring_core<char> can be backed by two representations. A “small” representation (a char array), and a “medium/large" representation (a char* + a size)
> I assume that the way you tell one from the other is
> 
> if (size == 0) small
> else medium-large
> 
> If my assumption is not correct, you’ll need to discover what the correct discriminator logic is - the class has to know, and so do you :-)
> 
> Armed with that knowledge, look in lldb source/Plugins/Language/CPlusPlus/Formatters/LibCxx.cpp
> There’s a bunch of code that deals with formatting llvm’s libc++ std::string - which follows a very similar logic to your class
> 
> ExtractLibcxxStringInfo() is the function that handles discovering which layout the string uses - where the data lives - and how much data there is
> 
> Once you have told yourself how much data there is (the size) and where it lives (array or pointer), LibcxxStringSummaryProvider() has the easy task - it sets up a StringPrinter, tells it how much data to print, where to get it from, and then delegates the StringPrinter to do the grunt work
> StringPrinter is a nifty little tool - it can handle generating summaries for different kinds of strings (UTF8? UTF16? we got it - is a \0 a terminator? what quote character would you like? …) - you point it at some data, set up a few options, and it will generate a printable representation for you - if your string type is doing anything out of the ordinary, let’s talk - I am definitely open to extending StringPrinter to handle even more magic
> 
> Part 2 - Teaching std::string that it can be backed by an fbstring_core
> 
> At the end of part 1, you’ll probably end up with a FBStringCoreSummaryProvider() - now you need to teach LLDB about it
> The obvious thing you could do would be to go in CPlusPlusLanguage::GetFormatters() add a LoadFBStringFormatter(g_category) to it - and then imitate - say - LoadLibCxxFormatters()
> 
>     AddCXXSummary(cpp_category_sp, lldb_private::formatters::FBStringCoreSummaryProvider, “fbstringcore summary provider", ConstString(“std::fbstring_core<.+>"), stl_summary_flags, true);
> 
> That will work - but what you would see is:
> 
>> (const string &const) corpus = error: summary string parsing error: {
>>   (std::fbstring_core<char>) store_ = “www"
> 
> You wanna do
> 
> (lldb) log enable lldb formatters
> (lldb) frame variable -T corpus
> 
> It will list one or more typenames - the most specific one is the one you like (e.g. for libc++ we get std::__1::string - this is how we tell ourselves this is the std::string from libc++)
> Once you find that typename, you’ll make a new formatter - FBStringSummaryProvider() - and register that formatter with that very specific typename
> 
> All that FBStringSummaryProvider() has to do is get the “store_” member (ValueObject::GetChildMemberWithName() is your friend) - and pass it down to FBStringCoreSummaryProvider()
> 
> 
> I understand this may seem a little convoluted and arcane at first - but feel free to ask more questions, and I’ll try to help out!
> 
>> Thanks.
>> Jeffrey
>> 
>> On Mon, Mar 28, 2016 at 11:38 AM, Enrico Granata <egranata at apple.com <mailto:egranata at apple.com>> wrote:
>> This is kind of orthogonal to your problem, but the reason why you are not seeing the kind of simplified printing Greg is suggesting, is because your std::string doesn’t look like any of the kinds we recognize
>> 
>> Specifically, LLDB data formatters work by matching against type names, and once they recognize a typename, then they try to inspect the variable in order to grab a summary
>> In your example, your std::string exposes a layout that we are not handling - hence we bail out of the formatter and we fall back to the raw view
>> 
>> If you want pretty printing to work, you’ll need to write a data formatter
>> 
>> There are a few avenues. The obvious easy one is to extend the existing std::string formatter to recognize your type’s internal layout.
>> If one were signing up for more infrastructure work, they could decide to try and detect shared library loads and load formatters that match with whatever libraries are being loaded.
>> 
>>> On Mar 28, 2016, at 9:47 AM, Greg Clayton via lldb-dev <lldb-dev at lists.llvm.org <mailto:lldb-dev at lists.llvm.org>> wrote:
>>> 
>>> So you need to be prepared to escape any text that can have special characters. A "std::string" or any container can contain special characters. If you are encoding stuff into JSON, you will either need to escape any special characters, or hex encode the string into ASCII hex bytes. 
>>> 
>>> In debuggers we often get bogus data because variables are not initialized, but the compiler tells us that a variable is valid in address range [0x1000-0x2000), but it actually is [0x1200-0x2000). If we read a variable in this case, a std::string might contain bogus data and the bytes might not make sense. So you always have to be prepared for bad data.
>>> 
>>> If we look at:
>>> 
>>>  store_ = {
>>>     = {
>>>      small_ = "www"
>>>      ml_ = (data_ =
>>> "��UH\x89�H�}�H\x8bE�]ÐUH\x89�H��H\x89}�H\x8bE�H\x89��~\xb4��\x90��UH\x89�SH\x83�H\x89}�H�u�H�E�H���\x9e���H\x8b\x18H\x8bE�H���O\xb4��H\x89ƿ\b",
>>> size_ = 0, capacity_ = 1441151880758558720)
>>>    }
>>>  }
>>> }
>>> 
>>> We can see the "size_" is zero, and capacity_ is 1441151880758558720 (which is 0x1400000000000000). "data_" seems to be some random pointer. 
>>> 
>>> On MacOSX, we have a special formatting code that displays std::string in CPlusPlusLanguage.cpp that gets installed in the LoadLibCxxFormatters() or LoadLibStdcppFormatters() functions with code like:
>>> 
>>>    lldb::TypeSummaryImplSP std_string_summary_sp(new CXXFunctionSummaryFormat(stl_summary_flags, lldb_private::formatters::LibcxxStringSummaryProvider, "std::string summary provider"));
>>>    cpp_category_sp->GetTypeSummariesContainer()->Add(ConstString("std::__1::string"), std_string_summary_sp);
>>> 
>>> Special flags are set on std::string to say "don't show children of this and just show a summary" So if a std::string contained "hello". So for the following code:
>>> 
>>> std::string h ("hello");
>>> 
>>> You should just see:
>>> 
>>> (lldb) fr var h
>>> (std::__1::string) h = "hello"
>>> 
>>> If you take a look at the normal value in the raw we see:
>>> 
>>> (lldb) fr var --raw h
>>> (std::__1::string) h = {
>>>  __r_ = {
>>>    std::__1::__libcpp_compressed_pair_imp<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >::__rep, std::__1::allocator<char>, 2> = {
>>>      __first_ = {
>>>         = {
>>>          __l = {
>>>            __cap_ = 122511465736202
>>>            __size_ = 0
>>>            __data_ = 0x0000000000000000
>>>          }
>>>          __s = {
>>>             = {
>>>              __size_ = '\n'
>>>              __lx = '\n'
>>>            }
>>>            __data_ = {
>>>              [0] = 'h'
>>>              [1] = 'e'
>>>              [2] = 'l'
>>>              [3] = 'l'
>>>              [4] = 'o'
>>>              [5] = '\0'
>>>              [6] = '\0'
>>>              [7] = '\0'
>>>              [8] = '\0'
>>>              [9] = '\0'
>>>              [10] = '\0'
>>>              [11] = '\0'
>>>              [12] = '\0'
>>>              [13] = '\0'
>>>              [14] = '\0'
>>>              [15] = '\0'
>>>              [16] = '\0'
>>>              [17] = '\0'
>>>              [18] = '\0'
>>>              [19] = '\0'
>>>              [20] = '\0'
>>>              [21] = '\0'
>>>              [22] = '\0'
>>>            }
>>>          }
>>>          __r = {
>>>            __words = {
>>>              [0] = 122511465736202
>>>              [1] = 0
>>>              [2] = 0
>>>            }
>>>          }
>>>        }
>>>      }
>>>    }
>>>  }
>>> }
>>> 
>>> So the main question is why are our "std::string" formatters not kicking in for you. That comes down to a typename match, or the format of the string isn't what the formatter is expecting.
>>> 
>>> But again, since you std::string can contain anything, you will need to escape any and all text that is encoded into JSON to ensure it doesn't contain anything JSON can't deal with.
>>> 
>>>> On Mar 27, 2016, at 9:20 PM, Jeffrey Tan via lldb-dev <lldb-dev at lists.llvm.org <mailto:lldb-dev at lists.llvm.org>> wrote:
>>>> 
>>>> Thanks Siva. All the DW_TAG_member related errors seems to go away after patching with your fix. The current problem is handling the decoding. 
>>>> 
>>>> Here is the correct decoding from gdb whic might be useful:
>>>> (gdb) p corpus
>>>> $3 = (const std::string &) @0x7fd133cfb888: {
>>>>  static npos = 18446744073709551615, store_ = {
>>>>    static kIsLittleEndian = <optimized out>,
>>>>    static kIsBigEndian = <optimized out>, {
>>>>      small_ = "www", '\000' <repeats 20 times>, "\024", ml_ = {
>>>>        data_ = 0x777777 <std::_Any_data::_M_access<void folly::fibers::Baton::waitFiber<folly::fibers::FirstArgOf<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::{lambda(folly::fibers::Promise<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::SelectionResult>)#1}, void>::type::value_type folly::fibers::await<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::{lambda(folly::fibers::Promise<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::SelectionResult>)#1}>(folly::fibers::FirstArgOf&&)::{lambda()#1}>(folly::fibers::FiberManager&, folly::fibers::FirstArgOf<folly::fibers::FirstArgOf<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::{lambda(folly::fibers::Promise<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::SelectionResult>)#1}, void>::type::value_type folly::fibers::await<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::{lambda(folly::fibers::Promise<facebook::servicerouter::RequestDispatcherBase<facebook::servicerouter::ThriftDispatcher>::prepareForSelection(facebook::servicerouter::DispatchContext&)::SelectionResult>)#1}>(folly::fibers::FirstArgOf&&)::{lambda()#1}, void>::type::value_type)::{lambda(folly::fibers::Fiber&)#1}*>() const+25> "\311\303UH\211\345H\211}\370H\213E\370]ÐUH\211\345H\203\354\020H\211}\370H\213E\370H\211\307\350~\264\312\377\220\311\303UH\211\345SH\203\354\030H\211}\350H\211u\340H\213E\340H\211\307\350\236\377\377\377H\213\030H\213E\350H\211\307\350O\264\312\377H\211ƿ\b", size_ = 0,
>>>>        capacity_ = 1441151880758558720}}}}
>>>> 
>>>> Utf-16 does not seem to decode it, while 'latin-1' does:
>>>>>>> '\xc9'.decode('utf-16')
>>>> Traceback (most recent call last):
>>>>  File "<stdin>", line 1, in <module>
>>>>  File "/mnt/gvfs/third-party2/python/55c1fd79d91c77c95932db31a4769919611c12bb/2.7.8/centos6-native/da39a3e/lib/python2.7/encodings/utf_16.py", line 16, in decode
>>>>    return codecs.utf_16_decode(input, errors, True)
>>>> UnicodeDecodeError: 'utf16' codec can't decode byte 0xc9 in position 0: truncated data
>>>>>>> '\xc9'.decode('latin-1')
>>>> u'\xc9'
>>>> 
>>>> Instead of guessing what kind of decoding I should use, I would use 'ensure_ascii=False' to prevent the crash for now.
>>>> 
>>>> I tried to reproduce this crash, but it seems that the crash might be related with some internal stl implementation we are using. I will see if I can narrow down to a small repro later. 
>>>> 
>>>> Thanks
>>>> Jeffrey
>>>> 
>>>> On Sun, Mar 27, 2016 at 2:49 PM, Siva Chandra <sivachandra at gmail.com <mailto:sivachandra at gmail.com>> wrote:
>>>> On Sat, Mar 26, 2016 at 11:58 PM, Jeffrey Tan <jeffrey.fudan at gmail.com <mailto:jeffrey.fudan at gmail.com>> wrote:
>>>>> Btw: after patching with Siva's fix http://reviews.llvm.org/D18008 <http://reviews.llvm.org/D18008>, the
>>>>> first field 'small_' is fixed, however the second field 'ml_' still emits
>>>>> garbage:
>>>>> 
>>>>> (lldb) fr v corpus
>>>>> (const string &const) corpus = error: summary string parsing error: {
>>>>>  store_ = {
>>>>>     = {
>>>>>      small_ = "www"
>>>>>      ml_ = (data_ =
>>>>> "��UH\x89�H�}�H\x8bE�]ÐUH\x89�H��H\x89}�H\x8bE�H\x89��~\xb4��\x90��UH\x89�SH\x83�H\x89}�H�u�H�E�H���\x9e���H\x8b\x18H\x8bE�H���O\xb4��H\x89ƿ\b",
>>>>> size_ = 0, capacity_ = 1441151880758558720)
>>>>>    }
>>>>>  }
>>>>> }
>>>> 
>>>> Do you still see the DW_TAG_member related error?
>>>> 
>>>> A wild (and really wild at that) guess: Is it utf16 data that is being
>>>> decoded as utf8?
>>>> 
>>>> As David Blaikie mentioned on the other thread, it would really help
>>>> if you provide us with a minimal example to repro this. Atleast, repro
>>>> instructions.
>>>> 
>>>> _______________________________________________
>>>> lldb-dev mailing list
>>>> lldb-dev at lists.llvm.org <mailto:lldb-dev at lists.llvm.org>
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev <http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev>
>>> 
>>> _______________________________________________
>>> lldb-dev mailing list
>>> lldb-dev at lists.llvm.org <mailto:lldb-dev at lists.llvm.org>
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev <http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev>
>> 
>> 
>> Thanks,
>> - Enrico
>> 📩 egranata@.com ☎️ 27683
>> 
>> 
> 
> 
> 
> Thanks,
> - Enrico
> 📩 egranata@.com ☎️ 27683
> 
> 


Thanks,
- Enrico
📩 egranata@.com ☎️ 27683

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20160407/ca3e0c16/attachment-0001.html>


More information about the lldb-dev mailing list