On Thu, Sep 12, 2024 at 10:27 AM David Marchand <david.march...@redhat.com> wrote: > > This is a continuation of the cleanup effort in logging macros. > > As a reminder of what this series is about, RTE_LOG_LINE() has been > introduced to check that the format string does not contain a trailing \n. > The goal was to prevent from: > - introducing multilines log messages (ugly and unhelpful for lambda > users), > - introducing double \n in logs, > - introducing missing \n in logs, > > I have built/maintained this series for fixing drivers/ since the > beginning of the year and I hope we can merge this series early in > 24.11. > > Some drivers are not converted because of multiple inconsistencies. > Those may be fixed later. > > As for the changes on lib/, new macros have been introduced so that > backports trigger a build error and force the backport to adjust the > patches containing logs accordingly. > > > Most of the changes have been done with some scripting, but it is likely > there are some errors and I hope driver maintainers will review in > depth. > > > A check is added for drivers/ in the hope that new drivers don't > introduce more mess. > > Yet, sometimes it is not possible to fix macros (a good example > is base/ drivers code...), so the check skips *osdep.h headers where > calls to RTE_LOG() may be accepted.
One note, there is an issue with the crypto_perf_cryptodev_perf test in (legacy) DTS. It (non explicitly) relied on the presence of an extra empty line, and the fact that a line would not contain special chars like : (what a joy..). I posted a fix for DTS. https://patchwork.dpdk.org/project/dts/patch/20240916092515.1851927-1-david.march...@redhat.com/ -- David Marchand
