This `select` clause had become very slow after we started caching
`ElementBase::toString` because the query used string concatenation to
produce alert messages, and those string concatenations were done very
early in the pipeline, producing lots of strings that would be discarded
moments later.
By using `$@` to interpolate elements into strings, the concatenation is
done outside of QL.
Testing on a Chromium snapshot, this commit takes us from
#select#ff ................ 6m2s
to
#select#cpe#134#fff ....... 15.2s
Logic is hard, and I made a mistake inverting the formula for the second case, so the
predicate never held for a sizeExpr like sizeof(int)*sizeof(void).
Now, this case is correctly handled by the fallback.
Based on PR feedback. This will avoid a syntactic wart and make the
invocation do the right thing both with and without
`language[monotonicAggregates]`.
This yields more precise size information in a lot of the common cases of C allocation code,
as the common pattern malloc(count * sizeof(type)) is now understood.
This predicate is effectively a Cartesian product between all enum
types. It's infeasible to compute it in full, so luckily the optimizer
has been able to apply enough magic to make it feasible. That's not a
robust solution, and it has indeed broken on at least one version of the
1.24 release candidate.
On a Chromium snapshot where I ran the LGTM suite overnight, the
`m#MistypedFunctionArguments::arithTypesMatch#bb` predicate (magic for
`arithTypesMatch`) took 170m5s. That was commit b69fdf5 from the
internal repo. I tried to reproduce it in VSCode, this time with commit
646646, but it wasn't quite as bad: the predicate took only 38 seconds.
In any case, making the problematic predicate `pragma[inline]` removes
the slow magic and makes the `MistypedFunctionArguments.ql` query
faster.
IR generation was not handling the special two-operand flavor of the `?:` operator that GCC supports as an extension. The extractor doesn't quite give us enough information to do this correctly (see github/codeql-c-extractor-team#67), but we can get pretty close.
About half of the code could be shared between the two-operand and three-operand flavors. The main differences for the two-operand flavor are:
1. The "then" operand isn't a child of the `ConditionalExpr`. Instead, we just reuse the original value of the "condition" operand, skipping any implicit cast to `bool` (see comment for rationale).
2. For the three-operand flavor, we generate the condition as control flow rather than the computation of a `bool` value, to avoid creating unnecessarily complicated branching. For the two-operand version, we just compute the value, since we have to reuse that value in the "then" branch anyway.
I've added IR tests for these new cases. I've also updated the expectations for `SignAnalysis.ql` based on the fix. @rdmarsh2, can you please double-check that these diffs look correct? I believe they do, but you're the range/sign analysis expert.
