there's a tool in Android that packs relative relocations. saves several MiB on the chromium webview .so. long term we'd like to see gold support it directly, but in the meantime it's a post-processing phase.
On Thu, Mar 12, 2015 at 2:44 PM, stephen Turner <[email protected]> wrote: > > On Thu, Mar 12, 2015 at 5:34 PM, enh <[email protected]> wrote: >> >> thanks. i did see the recent elftoolchain discussion here but didn't >> realize that they were API compatible with the existing elfutils. i've >> raised bugs internally for people to look at switching one of our >> build tools over, and to consider this when the time comes to write a >> perf that can build with clang and run on all three platforms. >> > > Maybe i missed it, who is switching over? im assuming your referring to a > distribution? > >> >> On Thu, Mar 12, 2015 at 6:09 AM, stephen Turner >> <[email protected]> wrote: >> > >> > On Mar 12, 2015 2:30 AM, "Daniel Cegiełka" <[email protected]> >> > wrote: >> >> >> >> Hi Elliott, >> >> >> >> 2015-03-12 7:09 GMT+01:00 enh <[email protected]>: >> >> > oh, i had a suggestion for the roadmap... perf(1). there are two or >> >> > three problems with perf, all of which are the kind of problems that >> >> > people on this list are likely to be interested in: >> >> > >> >> > 1. perf itself is GPL. >> >> > 2. perf uses elfutils (libelf), which is GPL. >> >> > 3. you can't compile elfutils with clang. (heavy use of nested >> >> > functions, a GCC extension clang doesn't support. the elfutils guys >> >> > seem to be vehemently against fixing that. but then these are people >> >> > who use strndupa...) >> >> >> >> point 2 and 3? >> >> >> >> https://github.com/elftoolchain/elftoolchain/tree/master/libelf >> >> >> >> Best regards, >> >> Daniel >> >> >> >> >> > >> > Not sure if its worth the mention but they recently had a discussion (or >> > get >> > trolled?) About the trademark/license of the name libelf as well. >> > However it >> > didnt appear that the person making the arguement was in any legal >> > position >> > to act nor was he representing someone who could. >> > >> > Its a quiet but responsive list should you have any questions. The >> > origional >> > main branch is on sf. The git hubs are apparently devs without commit >> > access >> > keeping up with patches etc. As the main sf branch appears to be slow to >> > update? Hopefully that will change. >> > >> > Stephen >> > >> >> > >> >> > point 3 is the one that's going to cause problems because it's one of >> >> > a very small number of projects in the entire tree that can't be >> >> > built >> >> > with clang, and it's the only thing in the tree where fixing it would >> >> > be a major change *and* the upstream owners are actively against the >> >> > change. >> >> > >> >> > i suspect at some point Android will end up writing a Java >> >> > implementation of perf (because most third party app developers use >> >> > Windows), but a native implementation that also worked on Mac OS and >> >> > WIndows might still be something Android could use directly in the >> >> > IDE. but even something Linux-only would still be useful for platform >> >> > developers, and could replace the existing implementation. >> >> > >> >> > (i see people here talk about big things like linkers and compilers >> >> > and can't help but think of the smaller things i don't currently have >> >> > a good option for :-) ) >> >> > >> >> > --elliott >> >> > _______________________________________________ >> >> > Toybox mailing list >> >> > [email protected] >> >> > http://lists.landley.net/listinfo.cgi/toybox-landley.net >> >> _______________________________________________ >> >> Toybox mailing list >> >> [email protected] >> >> http://lists.landley.net/listinfo.cgi/toybox-landley.net >> >> >> >> -- >> Elliott Hughes - http://who/enh - http://jessies.org/~enh/ >> Android native code/tools questions? Mail me/drop by/add me as a reviewer. > > _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
