> The patch seems to have a stray 'needs_devtmpfs' at line 304:
Sorry, it's not stray, just non-idiomatic. The intent was to avoid additional filesystem checks in the case where we already know the answer. It should probably have been written:
> Also, I noticed (gcc didn't ;-) that the mknod parameters have been inadvertently
> transposed. To be safe we should also specify the permissions for the device nodes
> we're creating (and possibly explicitly set umask).
Right - I've pushed an update to the branch for these three issues.
> The patch seems to have a stray 'needs_devtmpfs' at line 304:
Sorry, it's not stray, just non-idiomatic. The intent was to avoid additional filesystem checks in the case where we already know the answer. It should probably have been written:
if (stat ("/dev/ptmx", &statbuf) < 0 || !S_ISCHR( statbuf. st_mode)
|| major(statbuf. st_dev) != 5 || minor(statbuf. st_dev) != 2
|| stat("/dev/pts", &statbuf) < 0 || !S_ISDIR( statbuf. st_mode) )
needs_ devtmpfs = 1;
> Also, I noticed (gcc didn't ;-) that the mknod parameters have been inadvertently
> transposed. To be safe we should also specify the permissions for the device nodes
> we're creating (and possibly explicitly set umask).
Right - I've pushed an update to the branch for these three issues.