diff options
author | Zach van Rijn <me@zv.io> | 2024-12-10 07:32:52 -0600 |
---|---|---|
committer | Zach van Rijn <me@zv.io> | 2024-12-10 07:42:50 -0600 |
commit | d9f9f132bc10db7823631c5dff2986ae66ac922a (patch) | |
tree | a43506e7e45e382c9d86f3b4bcd5ae463cac36fd | |
parent | 2adc1a30b7e078fb8e132c3f3fd47f6d14b10fdf (diff) | |
download | bootstrap-d9f9f132bc10db7823631c5dff2986ae66ac922a.tar.gz bootstrap-d9f9f132bc10db7823631c5dff2986ae66ac922a.tar.bz2 bootstrap-d9f9f132bc10db7823631c5dff2986ae66ac922a.tar.xz bootstrap-d9f9f132bc10db7823631c5dff2986ae66ac922a.zip |
bootstrap: update mcmtools for non-interactive builds.
The goal of this project is to have a one-click bootstrap
script that will produce a tarball suitable for bootstrapping.
It must not hang or require interactivity.
The mcmtools 'bootstrap' script will hang in certain
cross-endian QEMU user environments, such as this project.
The hang occurs in the 'getifaddrs()' function, which only
used by curl's configure script to gate one unimportant thing.
Without this update (see upstream commit for more details),
one would need to kill the stuck conftest to the same end.
Upstream is the cleanest (but not correct) place to implement
this workaround. Other options were considered, such as to
watch for various processes or conditions, or to LD_PRELOAD,
etc. etc. and all seemed flaky.
The rootfs tarball images can now be created without killing
the stuck conftest. After this QEMU issue is resolved:
* https://gitlab.com/qemu-project/qemu/-/issues/2485
Then this project can update QEMU appropriately and upstream
mcmtools can subsequently remove the kludge.
-rwxr-xr-x | bootstrap | 2 |
1 files changed, 1 insertions, 1 deletions
@@ -410,7 +410,7 @@ if ! test -d "${MTOOLS}"/sys/emus/bin; then # FIXME: no hard code test -d bootstrap \ || git clone ${CHAINS}/bootstrap.git; cd bootstrap; - git checkout 4a1313497953f31a9ecfd3a2f82b24e92b845789; + git checkout 8a3ed8c226d7da1cee375d95c3b0232f7b5700ac; ## seed rootfs # |