On Monday 08 March 2021 18:44:58 Bin Meng wrote: > On Mon, Mar 8, 2021 at 6:11 PM Marek Behun <marek.be...@nic.cz> wrote: > > > > On Mon, 8 Mar 2021 17:16:03 +0800 > > Bin Meng <bmeng...@gmail.com> wrote: > > > > > Hi Marek, > > > > > > On Sun, Mar 7, 2021 at 12:26 PM Marek Behún <marek.be...@nic.cz> wrote: > > > > > > > > Currently we use incremental linking (ld -r) to link several object > > > > files from one directory into one built-in.o object file containing the > > > > linked code from that directory (and its subdirectories). > > > > > > > > Linux has, some time ago, moved to thin archives instead. > > > > > > > > Thin archives are archives (.a) that do not really contain the object > > > > files, only references to them. > > > > > > > > Using thin archives instead of incremental linking > > > > - saves disk space > > > > - apparently works better with dead code elimination > > > > - makes things easier for LTO > > > > > > > > The third point is the important one for us. With incremental linking > > > > there are several options how to do LTO, and that would unnecessarily > > > > complicate things. > > > > > > > > On the other hand, by using thin archives we can make (via the > > > > --whole-archive use flag) the final linking behave as if we passed all > > > > the object files from the archives to the linking program as arguments. > > > > > > I don't think --whole-archive is required for LTO to work. > > > > It is. Linking fails if it is not used, for example for > > nokia_rx51_defconfig. > > Could you investigate why? Is this due to missing marking some > variables/functions as __used?
See email Message-ID: <20210306210045.pn3vzuarh23poxuv@pali> for details. > > > > > Switching to --whole-archive should be made conditionally when LTO is on, > > > otherwise for targets that don't have > > > -ffunction-sections/data-sections/--gc-sections specified, it will > > > create unnecessary bloat. > > > > OK I will push into CI without this flag for non-LTO and if it passes > > all tests I shall remove this flag for non-LTO. > > > > Regards, > Bin
