On Tue, 28 Nov 2023 at 14:50, Sverdlin, Alexander <alexander.sverd...@siemens.com> wrote: > > In general, || in shell scripts should be avoided because it ignores > > the reason for the failure (ignoring some fails may be ok, ignoring > > all fails is not ok, as it will cause silent regressions). > > 20231111 will be problematic because rdfind will be required, but there is > no make target yet. My patch is at least falling back to "make install", > so it supports both 20231030 and versions after 20231111. It's not ignoring > the errors, falling back to "|| make install" will eventually fail the whole > build.
So once someone sends an update to 20231111, they should also backport the patch that introduces the new make target, and use that target. And if there's an update to an even newer version, then the new target can be used without any patching. But for now, I think nothing needs to be done? The problem with || is this: imagine that you have specified A || B, and you introduced A in addition to B, assuming some specific reason X that A can fail. What if it starts failing for some other reason Y? Then the fall-through to B may be valid, or it may not, and whatever had caused Y needs to be fixed some other way than executing B. This fall-through will happen quietly and no one is going to notice it, and that's where the regression is. Alex
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#191373): https://lists.openembedded.org/g/openembedded-core/message/191373 Mute This Topic: https://lists.openembedded.org/mt/102849939/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
