XPost: linux.debian.bugs.dist   
   From: sanvila@debian.org   
      
   > > Can you tell if this parallel mode is the default, and if not, what   
   > > other reasons for a race condition can be?   
   >   
   > I am not aware of any "parallel mode" here. The postinst calls dracut   
   > just with >&2 to redirect stdout to stderr.   
      
   I refer to this:   
      
       if [[ $parallel != "yes" ]]; then   
           for i in *; do   
               [[ -f $i/modules.dep ]] || [[ -f $i/modules.dep.bin ]] || continue   
               "$dracut_cmd" --kver="$i" "${dracut_args[@]}"   
               _rc=$?   
               if [[ $_rc -gt 0 ]]; then   
                   printf "%s\n" "dracut[F]: image generation failed for kernel   
   '$i'." >&2   
                   ((ret += _rc))   
               fi   
           done   
       else   
           for i in *; do   
               [[ -f $i/modules.dep ]] || [[ -f $i/modules.dep.bin ]] || continue   
   -->         "$dracut_cmd" --kver="$i" "${dracut_args[@]}" &   
           done   
      
   This "&" at the end is what draw my attention and made me to remember   
   the funny bug in "sumo" which I pointed out before.   
      
   > > I would be willing to build linux-signed-amd64 a lot of times using a   
   > > modified dracut package with whatever debug changes we could add to   
   > > see what's going on (even if they are simple "echo foo"), so I'm   
   > > open for suggestions about those potential changes.   
   >   
   > Thanks. Could you try to run a test with this patch applied to dracut:   
   > https://github.com/dracut-ng/dracut-ng/pull/2109/changes   
   >   
   > Plus add an "exit 1" after those 3 dinfo/dwarning calls to let dracut   
   > fail, because I expect dracut to take the happy path.   
      
   This first modified version fails a lot less, maybe because by adding those   
   additional lines we make the race condition to be less likely.   
      
   But I can't see any special thing in the logs (are they redirected somewhere?)   
      
   > Can you test with just[*] the stderr redirection to /dev/null removed?   
      
   This second modified version fails again quite often, but at least   
   we can see the error from 3cpio:   
      
   [...]   
   I: /vmlinuz is now a symlink to boot/vmlinuz-6.18.5+deb14-cloud-amd64   
   I: /initrd.img is now a symlink to boot/initrd.img-6.18.5+deb14-cloud-amd64   
   /etc/kernel/postinst.d/dracut:   
   dracut: Generating /boot/initrd.img-6.18.5+deb14-cloud-amd64   
      
   thread 'main' panicked at library/std/src/io/stdio.rs:1165:9:   
   failed printing to stdout: Broken pipe (os error 32)   
   note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace   
   /usr/bin/dracut: line 2772: cpio: command not found   
   dracut[F]: Creation of /boot/initrd.img-6.18.5+deb14-cloud-amd64 failed   
   [...]   
      
      
   [*] Note that after cloning from salsa I still have to revert the   
   change made by Bastian so that it depends again on 3cpio|cpio. The   
   first time I forgot about such little detail and was suspiciously   
   surprised that it did not fail at all...   
      
      
   I believe the fact that 3cpio is written in Rust is related to this issue.   
      
   I'm going to try this instead and will let you know how it goes:   
      
   --- a/dracut.sh   
   +++ b/dracut.sh   
   @@ -1483,7 +1483,7 @@ elif [[ -n $persistent_policy && ! -d "/de   
   /disk/${persistent_policy}" ]]; then   
    fi   
      
    CPIO=cpio   
   -if 3cpio --help | grep -q -- --create; then   
   +if 3cpio --help 2> /dev/null | cat | grep -q -- --create; then   
        CPIO=3cpio   
    fi   
      
   You can try this for the equivalent simplified example:   
      
   if 3cpio --help | cat | true; then true; fi   
      
   Thanks.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)