| Summary: | apt-get stops if synaptic is running during package install. | ||
|---|---|---|---|
| Product: | Sisyphus | Reporter: | Iakunin Andrei <iakuninaa> |
| Component: | apt | Assignee: | Ivan Zakharyaschev <imz> |
| Status: | NEW --- | QA Contact: | qa-sisyphus |
| Severity: | minor | ||
| Priority: | P5 | CC: | boyarsh, glebfm, imz, iv, ldv, placeholder, vt |
| Version: | unstable | ||
| Hardware: | x86_64 | ||
| OS: | Linux | ||
| Bug Depends on: | 40348 | ||
| Bug Blocks: | |||
|
Description
Iakunin Andrei
2022-04-20 11:13:03 MSK
Thanks for the report! (I hope someone some day will think about it and implement a solution.) A few more lock-related observations: * Now, concurrent install/remove actions can lead to broken state at the end -- https://bugzilla.altlinux.org/40348 * Waiting for locks before starting another instance of apt in Debian -- https://unix.stackexchange.com/q/242539/4319 I'm afraid that having just a lock for apt wouldn't solve a potential race between plain rpm and apt-or-whatever, as described in https://bugzilla.altlinux.org/40348 ; if a more tricky locking scheme is implemented in rpm (like the one suggested there), that would make "apt-get install" (or "rpm -U") at least not start the installation as long as synaptic holds a read-lock on the database. (Probably, that is the reason for the described behavior, but I might be wrong.) |