Hi Andreas, Andreas Enge <[email protected]> writes:
> Hello, > > the branch fails completely both on i686 and aarch64: > https://qa.guix.gnu.org/branch/javascript-team/package-changes?i686-linux-change=broken&i686-linux-change=still-failing&i686-linux-change=unknown-to-failing&i686-linux-change=new-failing > https://data.qa.guix.gnu.org/gnu/store/247sf3x36xbf3yf2abgdkcwzlvpvyi5s-libnode-24.16.0.drv > https://data.qa.guix.gnu.org/gnu/store/jyv7aad1sr9hp3ynacad98vm48h7qj8r-libnode-24.16.0.drv > https://data.qa.guix.gnu.org/gnu/store/fxvcxh9j978x48bwf1sfvlazci2ny7sz-node-24.16.0.drv > https://data.qa.guix.gnu.org/gnu/store/ihxc9w1p8g3h4kb9ilh0r2h34a4v4wha-node-24.16.0.drv Thanks for the ping [although all these links seem dead now]. > > Could you have a look? I've found some obscure corner of c++ (to me at least) regarding [temp.names] requiring a 'template' in some cases; I think this may be one of such cases. This hits only 32 bit targets, and happens to be fixed in a later version of v8 (not on purpose afaik). I'm not sure later revisions of v8 will be backported to this version of node-lts, so for now an additional "substitute*" will have to do. I've confirmed it compiles and run through all the tests via --system=i686-linux. I'm still looking into the aarch64 stuff, as that involves me setting up qemu-binfmt first which will probably also lead to extra (local) test failures due to the binfmt stuff messing with asserts wrt memory use via reading out RSS. If someone else has some bandwidth (or even better, an aarch64 machine to test on), that'd be appreciated. I'll be mostly AFK for the next 3 days, so can continue looking into this by then if things are not picked up by someone else. HTH, - Jelle
