AboutHaiku using garbage for version

[johndrinkwater]

Currently AboutHaiku shows "[]" characters in the version field.

[korli]

Could you check if it is fixed or still valid ?

[johndrinkwater]

Still valid as of rev16260, via Schmid's vmware 'factory'.

Having spoken to Steffano about it, it looks like non-beos compiled libraries are getting corrupted resources(?)

The only change has been the addition of (Revision X), so that it now shows "[][] (Revision 16260)[]"

[johndrinkwater]

Problems with version

[korli]

* Bug 130 has been marked as a duplicate of this bug. *

[axeld]

This seems to be a Linux build problem; I've reassigned it to Ingo.

[bonefish]

I recently had the same problem with an image I built. I tracked it down to that libbe.so was created with the node ID of a deleted node which had a BEOS:APP_VERSION attribute. Due to the way we emulate attributes under Linux libbe.so "inherited" the attributes of the former node. That's the reason why I recently tried to make use of extended attributes under Linux to remedy the problem, BTW. Anyway, I removed the attributes directory and all objects, did a full rebuild and the problem was gone.Now that I think of it, my former explanation doesn't make much sense, though. I had found that "setversion libbe.so -system ..." read the version_infos from the bad attribute and replaced only the system version_info (which is the correct behavior) leaving the bad app version_info finally used in AboutHaiku untouched (and writing them to both attribute and resources). What I apparently missed, is that before setversion, mimeset is invoked (not the other way around, as I apparently thought) and should already have transferred the correct app version_info from the resources to the attribute. So it actually shouldn't matter that the bad attribute existed beforehand.Mmh, not sure how to track that down, now that I can't reproduce it anymore. If anyone has clues...

[bonefish]

Apparently the combination of rc's writing only the app version_info, BAppFileInfo::GetVersionInfo() failing when reading not both infos, and the problem with the attribute emulation under Linux caused the problem.

Fixed in 16320.