Fix negative subsec for time_point #3261
Merged
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
vitaut
reviewed
Jan 10, 2023
include/fmt/chrono.h
Outdated
| val -= std::chrono::seconds(1); | ||
| } | ||
|
|
||
|
|
vitaut
reviewed
Jan 10, 2023
test/chrono-test.cc
Outdated
Comment on lines
888
to
891
| EXPECT_EQ("59.000", fmt::format("{:%S}", epoch - 4 * d)); | ||
| EXPECT_EQ("59.250", fmt::format("{:%S}", epoch - 3 * d)); | ||
| EXPECT_EQ("59.500", fmt::format("{:%S}", epoch - 2 * d)); | ||
| EXPECT_EQ("59.750", fmt::format("{:%S}", epoch - 1 * d)); |
There was a problem hiding this comment.
I'm not sure if we need all four of these checks because they are effectively testing the same thing. One is probably enough.
vitaut
reviewed
Jan 10, 2023
test/chrono-test.cc
Outdated
Comment on lines
893
to
896
| EXPECT_EQ("00.250", fmt::format("{:%S}", epoch + 1 * d)); | ||
| EXPECT_EQ("00.500", fmt::format("{:%S}", epoch + 2 * d)); | ||
| EXPECT_EQ("00.750", fmt::format("{:%S}", epoch + 3 * d)); | ||
| EXPECT_EQ("01.000", fmt::format("{:%S}", epoch + 4 * d)); |
vitaut
reviewed
Jan 10, 2023
|
Thank you! |
|
This change triggers a ubsan error on the following case: #include <fmt/chrono.h>
int main() {
auto d = std::chrono::system_clock::duration(-9223372036854767584);
fmt::print("{:}", std::chrono::system_clock::time_point(d));
}Stack trace: |
|
I attempted to resolve this ub by first converting the #include <fmt/chrono.h>
int main() {
auto val = -9223372036854767584 / 1'000'000 - 1; // -9223372036855
auto tp =
std::chrono::time_point<std::chrono::system_clock, std::chrono::seconds>(
std::chrono::seconds(val));
auto time = std::chrono::system_clock::to_time_t(tp);
fmt::print("{} {}", val, time);
// expected: -9223372036855 -9223372036855
// actual: -9223372036855 9223372036854
}The message from the sanitizer is as follows: I don't know why the library would multiply the seconds by 1,000,000 and then convert it to |
|
An alternative solution is to subtract 1 from the resulting #include <fmt/chrono.h>
int main() {
auto val = -9223372036854; // -9223372036854767584 / 1'000'000
auto tp =
std::chrono::time_point<std::chrono::system_clock, std::chrono::seconds>(
std::chrono::seconds(val));
// This results in integer overflow since to_time_t does some multiplications.
auto t1 = std::chrono::system_clock::to_time_t(tp - std::chrono::seconds(1));
// This assumes time_t to represent seconds and is signed, which is not
// specified by the standard, but true for almost all platforms.
auto t2 = std::chrono::system_clock::to_time_t(tp) - 1;
fmt::print("t1: {}, t2: {}\n", t1, t2);
// t1: 9223372036854, t2: -9223372036855
} |
|
I turned the UB into an error in 240b728. If you have a better idea, a PR would be welcome. |
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.
Fix #3117
The following piece of code (https://godbolt.org/z/4qr5389ns) used to print
59.000 00.750 00.500 00.250 00.000 00.250 00.500 00.750 01.000, which is wrong. After this PR, it now correctly prints59.000 59.250 59.500 59.750 00.000 00.250 00.500 00.750 01.000