Public bug reported:
Problem reports are deb822 style text files. The CoreDump is zstd or
gzip compressed and base64-encoded. This add an overhead of 1/3 to the
on-disk size. In most cases the CoreDump dominates the size of the crash
report, making the overhead noticeable.
Consider using binary format like cbor/msgpack to store/submit the
report. I wrote a simple POC to convert the report:
```
#!/usr/bin/python3
import argparse
import pathlib
import problem_report
import cbor2
DEFAULT_SUFFIX = ".cbor"
def write_output(report: dict[str, str | bytes], output: pathlib.Path) -> None:
with output.open("wb") as f:
cbor2.dump(report, f)
def load_report(input: pathlib.Path, decompress: bool) -> dict[str, str |
bytes]:
report = problem_report.ProblemReport()
with input.open("rb") as f:
report.load(f, binary=True if decompress else "compressed")
return report.as_dict()
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("-d", "--decompress", action="store_true")
parser.add_argument("-s", "--suffix", default=DEFAULT_SUFFIX)
parser.add_argument("input", type=pathlib.Path)
args = parser.parse_args()
report = load_report(args.input, args.decompress)
output = pathlib.Path(f"{args.input}{args.suffix}")
print(f"Using output file: {output}")
write_output(report, output)
if __name__ == "__main__":
main()
```
Example size reduction:
371M _usr_bin_gitg.1000.crash
279M _usr_bin_gitg.1000.crash.cbor
11M _usr_libexec_localsearch-3.1000.crash
7.9M _usr_libexec_localsearch-3.1000.crash.cbor
** Affects: apport (Ubuntu)
Importance: Wishlist
Status: New
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2164750
Title:
Consider using a binary format like cbor/msgpack for crash reports
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/apport/+bug/2164750/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
