-Luis will try to dive to look into this August 19th - Sept 13th. With a simple proof of concept feature / driver.

-

-= SmPL integration =

-

-Typically patches may fail if hunk offsets are off by the allowed offset. We upkeep hunk offsets placements up to date by refreshing patches. We used to use quit refresh to refresh patch offsets but today we simply extract the new delta after application of a patch in which hunk offsets were off. At times however patches can fail to apply due to contexual changes within the patch. We are considering using SmPL to help with this.

-

-Its plausible that if you're hacking on the Linux kernel that you may have not yet heard of [http://coccinelle.lip6.fr/ coccinelle] or [http://coccinelle.lip6.fr/sp.php SmPL patches]. Truth is SmPL is already widely used across many subsystem in the Linux kernel and SmPL's usage even extends outside of the Linux kernel. The page that tries to document the [http://coccinelle.lip6.fr/impact_linux.php impact of SmPL on the Linux kernel] is clearly out of date and likely just needs to be automated with ''git grep''. The idea behind the need for SmPL came about studying and formalizing the concept of [http://coccinelle.lip6.fr/ce.php collateral evolutions] on the Linux kernel. Luis elaborates on his blog on how this can be used to help [http://www.do-not-panic.com/2012/08/optimizing-backporting-collateral.html optimize collateral evolutions]. While still theoretical, if we want to make backporting more efficient we need to start using SmPL soon. The work should consist of properly supplying expressions that summarize backporting collateral evolutions from patches from which SmPL could likely deduced or written manually. In order accomplish this each collateral evolution must first be atomically reduced as much as possible. We then will need to categorize backported collateral evolution and prioritize working on categories that we could likely infer SmPL from. As it stands patches which have been already reduced as much as possible have been annotated by using a four digit prefix. Patches which can likely be split up further remain with a two digit prefix.

-

-Patches that need to be reviewed are under the [https://git.kernel.org/cgit/linux/kernel/git/mcgrof/backports.git/tree/patches/collateral-evolutions patches/collateral-evolutions] directory. The way backports works -- any directory name can be used to separate patches into any logical form as required. Integration with SmPL can start with inferring SmPL from the few four-digit prefixed patches and subsequently writing a Python tool which would use the SmPL patch to patch kernel code. The first objective is to prove backporting transformations are equivalent by either using the old series of patches for a specific collateral evolution or by using its expression in SmPL form. A proof can be obtained by computing the sha256 delta on the tree with both approaches.

-

-== Who's working on this? ==

-

-Luis will try to dive to look into this August 19th - Sept 13th. With a simple proof of concept feature / driver. Luis hopes to further collaborate with Julia on this between October 16 - December 15.

-

-= Infer SmPL for patch hunks healing =

-

-If a collateral evolution does not yet have a formal SmPL expression to represent it and if SmPL inference is plausible it'd be great to infer SmPL on the fly as a patch fails to attempt fixing the patch. SmPL inference however is only possible today with a series of similar hunks, at least 2, across different set of offsets. SmPL hunk healing then should not be possible for addressing singular hunk collateral evolutions.

-

-== Who's working on this? ==

-

-Luis hopes to further collaborate with Julia on this between October 16 - December 15.

-

-= Using the inverse of SmPL to backport =

-

-If Linux kernel development dealt with collateral evolutions with a strict SmPL form it is theoretically possible that one might be able to use the inverse of a patch to backport a collateral evolution. We need to prove / disprove this. Exceptions to this and corner cases should also be inspected.

-

-== Who's working on this? ==

-

-Luis hopes to further collaborate with Julia on this between October 16 - December 15.