You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@arrow.apache.org by "Alessandro Molina (Jira)" <ji...@apache.org> on 2021/05/20 14:30:00 UTC
[jira] [Commented] (ARROW-12837) [C++] Array::ToString() segfaults
with null buffer.
[ https://issues.apache.org/jira/browse/ARROW-12837?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17348561#comment-17348561 ]
Alessandro Molina commented on ARROW-12837:
-------------------------------------------
I guess it might be reasonable to check {{array.data_.length}} in {{ArrayPrinter}} to print some other value instead of the content when the array is made of empty data. Also checking {{raw_values != nullptr}} might be another easy check.
> [C++] Array::ToString() segfaults with null buffer.
> ---------------------------------------------------
>
> Key: ARROW-12837
> URL: https://issues.apache.org/jira/browse/ARROW-12837
> Project: Apache Arrow
> Issue Type: Bug
> Components: C++
> Affects Versions: 4.0.0
> Reporter: Alessandro Molina
> Assignee: Alessandro Molina
> Priority: Minor
>
> In some tests in the Arrow test suite there are cases where arrays are made without allocating real buffers of data, using {{Buffer}} of length 0 and with {{nullptr}} as their data.
> In this cases the {{Array::ToString()}} does lead to segfault, at least for Int arrays.
> {code}
> * thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
> frame #0: 0x0000000101dac696 libarrow.500.dylib`std::__1::enable_if<is_integer_type<arrow::NumericArray<arrow::Int32Type>::TypeClass>::value, arrow::Status>::type arrow::ArrayPrinter::WriteDataValues<arrow::NumericArray<arrow::Int32Type> >(this=0x00007ffeefbfea98, i=0)::'lambda'(long long)::operator()(long long) const at pretty_print.cc:152:73
> 149 enable_if_integer<typename T::TypeClass, Status> WriteDataValues(const T& array) {
> 150 const auto data = array.raw_values();
> 151 // Need to upcast integers to avoid selecting operator<<(char)
> -> 152 WriteValues(array, [&](int64_t i) { (*sink_) << internal::UpcastInt(data[i]); });
> 153 return Status::OK();
> 154 }
> 155
> {code}
> The issue can probably be worked around by checking if the data in the array is valid (at least for the most obvious cases, like NULLS or 0-length data) when building the string representation
--
This message was sent by Atlassian Jira
(v8.3.4#803005)