cmake: enable Wextra #2542
cmake: enable Wextra #2542
Conversation
Riolku
force-pushed
the
enable-wextra
branch
4 times, most recently
from
cce53ae
to
040902f
Compare
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #2542 +/- ##
=======================================
Coverage 92.85% 92.85%
=======================================
Files 1025 1025
Lines 38338 38323 -15
=======================================
- Hits 35598 35585 -13
+ Misses 2740 2738 -2 ☔ View full report in Codecov by Sentry. |
src/common/arrow/arrow_row_batch.cpp
Outdated
| @@ -256,7 +256,7 @@ void ArrowRowBatch::templateCopyNonNullValue<LogicalTypeID::VAR_LIST>( | |||
| auto numBytesForChildValidity = getNumBytesForBits(numChildElements); | |||
| vector->childData[0]->validity.resize(numBytesForChildValidity); | |||
| // Initialize validity mask which is used to mark each value is valid (non-null) or not (null). | |||
| for (auto i = currentNumBytesForChildValidity; i < numBytesForChildValidity; i++) { | |||
| for (int64_t i = currentNumBytesForChildValidity; i < numBytesForChildValidity; i++) { | |||
There was a problem hiding this comment.
I am a little confused here. What is the warning message we got before? I think numBytesForChildValidity is a uint type, why do we make i as int64_t rather than uint64_t?
There was a problem hiding this comment.
numBytesForChildValidity is a int64_t.
| @@ -104,8 +104,9 @@ void PyConnection::getAllEdgesForTorchGeometric(py::array_t<int64_t>& npArray, | |||
| auto dstBuffer = buffer + numRels; | |||
| for (int64_t batch = 0; batch < batches; ++batch) { | |||
There was a problem hiding this comment.
If the batch number must be a non-negative number, should we make it an unsigned int?
There was a problem hiding this comment.
Probably - I've made a small edit to address this, and added a comment for why start and end are int64_t.
Primarily these warnings fall into two categories: - Unused parameter (that clang-tidy didn't catch). - Mismatch sign comparison. Generally this happens in loops, and the correct fix is `auto i = 0u`. This gives `i` the type of `unsigned`, and so we must be careful that `i` cannot overflow 2^32. Wextra is disabled when building third_party code.
Primarily these warnings fall into two categories:
auto i = 0u. This givesithe type ofunsigned, and so we must be careful thaticannot overflow 2^32.Wextra is disabled when building third_party code.