Control: forwarded -1 https://marc.info/?l=linux-aio&m=171745321208314
Hi! On Sun, 2024-06-02 at 14:28:31 +0000, Romain Geissler wrote: > Package: libaio1t64 > Version: 0.3.113-8 > Severity: normal > When installing the "new" package libaio1t64 package in Debian testing > or Ubuntu Noble, the previously available libaio.so.1 file/symlink is no > longer available. This effectively breaks non-debian built package like > for example the Oracle Instant client which would need a libaio.so.1 > compat symlink to "work". This was for example reported in Oracle forums > here: > https://forums.oracle.com/ords/r/apexds/community/q?question=instant-client-on-ubuntu-24-04-noble-numbat-7244 > I see that for #1067831 you added a compat symlink, but only for the > udeb package, and apparently in the dicussion the idea to add this > symlink in the base package rather the udeb one was ruled out. However > in practice it means third party binaries like the oracle ones are still > broken unless a symlink is manually created. In the Debian context this was intended to be a temporary non-intrusive solution that would not stomp over upstream, until this had been agreed with them. My intention has always been to revert the local SONAME bump before the next Debian release. Ubuntu will have to deal with this however best they see fit. > If there is really an issue with adding it in the package, can we at > least consider Raphael Hertzog's propsal to add it as a postinstall > step (checking if the libaio.so.1 exists and if not symlink it) ? I was able to verify whether part of the Linux kernel fix I had in mind solved the test suite issues (and were not caused by faulty libaio changes), and have now submitted the libaio and Linux fixes upstream, at: https://marc.info/?l=linux-aio&m=171745321208314 so ideally, if they agree with the API/ABI changes, I'd then revert the SONAME transition in Debian and provide the t64 backwards compat symlinks in the libaio1 package. Otherwise I'll work with upstream on a suitable API/ABI change, and in that case I would not be able to provide backwards compat symlinks. Thanks, Guillem