Fix many compiler warnings #831
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The compiler complains about comparisons between int and unsigned int in a handful of cases in these two files. The integers are being used for loops, and it seems clear that unsigned int is a fine replacement for signed in these cases.
This file produces many warnings of the form
```
tensorflow_quantum/core/src/circuit_parser_qsim.cc:185:8:
warning: variable 'unused' set but not used [-Wunused-but-set-variable]
185 | bool unused = absl::SimpleAtoi(op.qubits(0).id(), &q0);
| ^
```
Inspecting the code reveals that variable `unused` is indeed set but
never used. I replaced the code with simple testing of the return
values, similar to what is used in one place in the file, and in a way
that follows the recommendations from the Abseil docs.
One of the cases of lambda capture in this file resulted in warnings
of this sort:
```
tensorflow_quantum/core/ops/tfq_simulate_state_op.cc:154:29:
warning: lambda capture 'max_num_qubits' is not used [-Wunused-lambda-capture]
154 | auto copy_f = [i, nq, max_num_qubits, &output_tensor, &ss, &sv](
| ~~^~~~~~~~~~~~~~
```
The parameter really was not used in the body. Removing it from the
lambda capture silenced the warning.
Previously, util_qsim.h included the definitions of the function `BalanceTrajectory`. This led to compiler warnings about undefined functions in places that included util_qsim.h but didn't use all variants of `BalanceTrajectory`. This refactor moves `BalanceTrajectory` to its own file (util_balance_trajectory.cc), and also moves the relevant tests from util_qsim_test.cc to the corresponding util_balance_trajectory.cc.
`INSTANTIATE_TEST_CASE_P` is deprecated and produces compilation warnings. The replacement is `INSTANTIATE_TEST_SUITE_P`.
A number of harmless warnings during compilation come from third-party things pulled in by the TFQ build rules. They involve things like unused functions, unused variables, or (in the case of zlib) some deprecated uses of prototypes. The best way to fix them would be in the original source code, but in some cases this is impossible; for example, we can't change what TensorFlow 2.15.0 needs. The second best way would be to add patch rules in TFQ to adjust the code or the build rules. I spent a couple of hours trying to figure out how to that for the copy of zlib included by bazel_rules, but due to my still-limited Bazel knowledge, I wasn't able to get the patching scheme working. So, for the time being, and in consideration of the harmless nature of the warnings, the approach taken here is to add compile-time per-file flags that tell the compiler not to warn about the specific kinds of problems that we know exist in the files involved.
Many files produced many type mismatch warnings with both clang 17 and gcc 14 on a Debian 6.10.11 x86_64 system. This commit fixes almost all the ones I'm seeing in the TFQ code base. Some other warnings still arise in other code and modules. Those warnings don't appear to be things we can easily work around short of creating a lot of patch files (at significant cost in time and testing). For those, I'm currently relying on setting compiler options (e.g., `-Wno-unused-function` for warnings about unused functions) narrowed in scope to the problematic package using Bazel's `--per_file_copt` and `--host_per_file_copt` flags. Those flags are not in this commit; they need to be made part of the TFQ `configure.sh` script.
mhucka
added
the
area/code-health
In more recent versions of Python like 3.10, the `base64` package changed APIs, and this in turn caused nbformat 4.4.0 to fail to work. Updating to the latest version (5.1.3) solved that problem. A minor tweak was also needed in scripts/format_ipynb.py to account for a different error class being returned when "!pip ..." commands are encountered in the notebook file.
After updating nbformat, running scripts/format_ipynb.py changed the contents of all the ipython notebook files. Most frustratingly, one of the changes is that the indentation of the .ipynb files changed, making the diffs particularly noisy. Spot-inspecting files manually reveals a number of real changes, some being worthwhile (e.g., adding missing spaces in argument lists) and some being just differences in formatting such as where line breaks are introduced. The latter set of changes is puzzling because the style settings haven't changed. I tried but couldn't find a way to avoid these changes except by outright changing the style parameters, but then, that would mean the TFQ .ipynb file formatting would become non-standard, and that seems worse. So, for lack of a better solution, I'm checking in all the reformatting notebooks with the hope that future versions of yapf and nbformat don't keep introducing more .ipynb format changes.
Forgot to run the formatter.
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
MichaelBroughton
requested changes
Dec 12, 2024
There was a problem hiding this comment.
This is pretty big and contains a lot of changes that don't have a lot to do with each other. Could we split this into a few smaller PRs ? Perhaps something like:
- The notebook format as it's own PR.
- The C++ type changes and some warning fixes.
- The refactoring of trajectory balancing.
- The introduction of the additional error checking in some of the parse code.
As it stands there is a lot going on in each of these areas and I worry if we miss things and require a rollback, we also won't have a commit history that lets us separate out things easily.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This fixes a large number of Bazel build and compiler warnings. Summary:
int's being used where unsigned int or vector size types should be used instead, to match the type of the value the loop variable was being compared against.tensorflow_quantum/core/src/circuit_parser_qsim.cc. These involved status return values from Abseil function calls. I replaced the code with simple testing of the returnvalues, similar to how it was already done in one place in the file, and in a way that follows the recommendations from the Abseil docs.
BalanceTrajectoryfromutil_qsim.hinto a separate file (util_balance_trajectory.cc), and make adjustments in the BUILD file and files that used the function. This was to eliminate repeated warnings about unused functionBalanceTrajectoryin files that includedutil_qsim.h. Also moved the relevant tests fromutil_qsim_test.ccto their own file.configure.shto silence warnings we can't do anything about because they're coming from packages included as a result of importing TensorFlow Bazel workspace files. The options are scoped as narrowly as possible to limit the impact of hiding warnings to specific cases.I tested these changes with both gcc 14.2.0 and clang 17.0.6 on a Debian 6.10 Linux system.