quickdump v2: relative key names, reused buffer, etc.

[kodebach]

A small PR with some changes to quickdump. We now only store relative key names, to enable relocation. We also reuse the buffer on kdbGet, for a minor speed-up.

@markus2330 please review my pull request

[markus2330]

Thank you for the fix! It is much better to have relative key names, see also #51.

[kodebach]

From https://travis-ci.org/ElektraInitiative/libelektra/jobs/507354465#L1741:

No plugin that provides sync could be found

@markus2330 How should I fix that? The sync plugin is excluded for the Travis GCC ASAN job (I don't know why), should I re-include it or should I exclude quickdump? (locally GCC ASAN works, if the sync plugin is there) Or is there something I can do, so that sync isn't required?

[markus2330]

Sync is one of the essential plugins which should always be present. In #423 we discussed about how to improve without risking data loss. So at the moment it would be best if the sync plugin is included in every build job.

@sanssecours Do you know why the sync plugin is not included?

[sanssecours]

Do you know why the sync plugin is not included?

Yes. I excluded sync and all other non-essential plugins to improve the runtime of 🐧 GCC ASAN.

[kodebach]

Yes. I excluded sync and all other non-essential plugins to improve the runtime of 🐧 GCC ASAN.

If we want to only include essential plugins in 🐧 GCC ASAN shouldn't we use:

-DPLUGINS='dump;fstab;haskell;jni;journald;list;lua;python;python2;regexdispatcher;resolver;ruby;simpleini;spec;typechecker;xerces'

(the list of plugins available at the time of your commit which you didn't exclude)

Instead of starting with ALL and excluding a whole bunch of plugins?

[sanssecours]

…Instead of starting with ALL and excluding a whole bunch of plugins?

I explicitly excluded the old (and tested) plugins, since I wanted to make sure newly added plugins (that do not depend on any of the excluded plugins) are tested by 🐧 GCC ASAN.

[markus2330]

Thank you!