Harald van Dijk wrote in <[email protected]>: |On 06/09/2022 20:39, Steffen Nurpmeso wrote: |>|What was the initial expression that did not work properly with the old |>|implementation? |> |> Well if you run the test of the patchset against busybox with the |> old implementation you will see multiple errors (of course >>>[=] |> and **= because they are not supported), mostly ?: related. To |> make it plain, ?: is broken but for simple cases. | |Relying on readers to run your test suite that contains lots of tests |for functionality that busybox does not currently aim to implement, and
No, really now. And the rest is also not right. You can see in existing tests that lot is commented out because it does not work. It does not deal with whiteouts, as i initially also did wrong, for example. For example variables are expanded in there, which is wrong. Also if you go in recursion there ... but really, run the test, remove the five or six >>>=? and **= cases, and you will see that many, many ?: tests will fail. That Dijkstra algorithm really scales badly for anything beyond simple binary, and the solution is not as easy as what the current implementation has. ... |The supplied patch appears to fix both #1 and #2, but it has alignment |issues: when UBSAN is enabled, I see a lot of "runtime error: store to |misaligned address <...> for type 'char *', which requires 4 byte |alignment" warnings when using shell arithmetic. ASAN works, UBSAN i have not tried. How can char* require 4 byte alignment? --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt) _______________________________________________ busybox mailing list [email protected] http://lists.busybox.net/mailman/listinfo/busybox
