94xhn opened a new pull request, #722:
URL: https://github.com/apache/logging-log4cxx/pull/722
## Problem
`MDCPatternConverter::format()` computes the remaining space budget for the
"full MDC" JSON output as:
```cpp
size_t remainingLength = info.getMaxLength() - 1;
```
`getMaxLength()` returns an `int`. When a pattern specifies a maximum field
width of zero (e.g. `%.0J`, or `%mdc` configured with precision `0`), this
evaluates as `0 - 1` on an unsigned `size_t`, wrapping around to `SIZE_MAX`.
The subsequent guard `if (remainingLength < itemLength) break;` then never
fires, so the truncation is silently defeated and the **entire** MDC map is
written out instead of nothing — the opposite of what the pattern requested.
This is reachable through any normal `PatternLayout` conversion pattern
using `%J`/`%.NJ` (registered via `RULES_PUT("J", MDCPatternConverter)` in
`patternlayout.cpp`), as well as through `%mdc`. `patternparser.cpp`'s
`DOT_STATE` stores the digits after `.` directly as `maxLength` with no
lower-bound check, so `maxLength == 0` is a legitimate, reachable value, not a
theoretical one.
The generic truncation path used elsewhere (`FormattingInfo::adjustField()`)
already handles `maxLength == 0` safely (plain cast, no subtraction), so this
"full MDC" branch's own hand-rolled counter is the actual outlier/bug, not an
intentional deviation.
## Fix
Clamp before subtracting, matching the safe-cast pattern already used in
`FormattingInfo::adjustField()`:
```cpp
size_t remainingLength = info.getMaxLength() > 0
? static_cast<size_t>(info.getMaxLength()) - 1
: 0;
```
## Testing
Added `MDCTestCase::test3`, which exercises `%.0J` against an MDC containing
one entry and asserts the output now correctly contains no MDC content, instead
of the full (unbounded) map. Existing `test2` (`%.30J`) and default/`INT_MAX`
cases are unaffected — verified the arithmetic is unchanged for all `maxLength
> 0` values, only the `maxLength == 0` case changes (from wrap-around to
correctly-clamped `0`).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
