Hi, "Michael F. Lamb" <mike@orbital.rodeo> skribis:
> I was following the instructions in the Guix Reference Manual for running > Guix in a VM using the pre-built qcow2 VM image: > > https://guix.gnu.org/manual/en/guix.html#Running-Guix-in-a-VM > > The documentation instructs me to create a qemu VM with the option '-m 1024' > which provides it 1GB of RAM. > > After doing so, each time I attempted to run 'guix pull', I received the > error message: > >> You found a bug: the program '/gnu/store/...-compute-guix-derivation' failed >> to compute the derivation for Guix (version: "..."; system: "x86_64-linux"; >> host version: "..."; pull-version: 1). > > Searching for this error message led me to many reports where "just run 'guix > pull' again" eventually worked for the reporter but such was not the case for > me. > > Watching "top" while running "guix pull" showed the memory usage increasing > to 100% whereupon "guix pull" fails. I set the -m option to '4096' and > thereafter 'guix pull' worked for me. > > A few approaches you might take: > > 1. Make 'compute-guix-derivation' or the process that executes it better at > reporting what variety of failure has occurred. > 2. Change the docs to increase the default amount of memory granted to the > VM. (But this might not be helpful for users with older machines and limited > available memory.) > 3. Change the docs to provide the VM with swap space. > 4. Attempt to reduce the amount of memory compute-guix-derivation requires to > complete. It looks like the memory requirements to build the latest revisions of Guix have increased (and this is a bit ridiculous). I checked with <https://ftp.gnu.org/gnu/guix/guix-system-vm-image-1.3.0.x86_64-linux.qcow2> and ‘-m 2048’ gives us enough headroom, so I modified the manual accordingly in commit 98a8b48a69b8208475c9a1e40d09517f8643b8cb. Thanks for your report! Ludo’.