only the author date was being set before
with this change, at least on my system, i
now get the same commit ID and date each time
on the following test:
NOTE: . is lbmk (main work directory)
NOTE: here, we are copying the current clone
git clone . lbmktest
cd lbmktest
./mk version
rm -Rf .git
./mk version
git log
same output each time. we have thus eliminated
timestamps as a variable. this is but a part of
in a series of changes that will eventually lead
to libreboot being built reproducibly.
Signed-off-by: Leah Rowe <leah@libreboot.org>
don't even call the date util at all, to fix the
portability issues. now we don't need GNU date
anymore.
git itself can just take an epoch, and it will
do whatever it will do. yes.
Signed-off-by: Leah Rowe <leah@libreboot.org>
even if i unset PATH before running lbmk, it is
reinitialised because of the shebang. if no PATH
exists, then it's either a broken system and you
have bigger problems, or you have some weird
system that uses something else.
our job is not to fix the world. there was no
scenario in which this check would even trigger
an error or fallback, so just remove it.
Signed-off-by: Leah Rowe <leah@libreboot.org>
i've been careful to makely only use it on AND operations,
not onse that use OR. a lot of blocks are not condensed,
unlike previously when this design was used.
i removed the condensed design because it made the code
allegedly easier to read, but i found it harder to read
and found the code looked dirty. this change makes it
clean again, but i've done it in a way where the shorthand
conditional statements are easy to understand for most
people. this strikes a compromise; i would go further.
Signed-off-by: Leah Rowe <leah@libreboot.org>
use the simpler logic recently introduced, generalised
but it only sets variables to empty. this is safer than
the previous design, and brings the same reduction in
sloccount overall since i wasn't doing anything other
than emptying variables anyway, when it was used before.
i didn't like the long list of variables inside files,
so this makes everything more readable.
Signed-off-by: Leah Rowe <leah@libreboot.org>
not needed at the moment, since we ensure that
the functions always exit with error under fault
condition. they always return zero.
nevertheless, this might catch some unexpected
behaviour in the future.
Signed-off-by: Leah Rowe <leah@libreboot.org>
we don't need a main here. it's so generic that we
can just leave it really small like it is now. now
it's much clearer what it actually does.
Signed-off-by: Leah Rowe <leah@libreboot.org>
commands now fed from a file, include/common.sh
which in turn also includes other lib files
this makes mk completely generic, and means
that changes will now be completely in sync
with cbmk, ensuring a lack of merge conflicts,
though the file was already pretty small anyway.
Signed-off-by: Leah Rowe <leah@libreboot.org>
i made several modificatinos to several files
thus for in 2026, in the main part of the build system.
i've added 2026 to the ones that i modified.
Signed-off-by: Leah Rowe <leah@libreboot.org>
now you can put lbmk somewhere on your system,
and symlink to the mk script from PATH
e.g. /usr/local/bin/lbmk -> /home/leah/lbmk/mk
lbmk will cd to itself, including from PATH,
because the string will then be passed through
readlink, which gets the real path.
Signed-off-by: Leah Rowe <leah@libreboot.org>
i need only now implement a check for sys dirs
e.g. /etc/lbmk.cfg and check if it defines
e.g. other location for config/ and include/
then mk is a generic dispatcher and i can one
day make mk possible to install. for example, if
you have a bsd ports system and you want to
install lbmk
i could make it check user dirs too, overriding
defaults. for example /usr/local/etc/lbmk, but
then i might have ~/.config/lbmk
you no longer need to be in the xbmk directory
to run xbmk. it'll cd to itself, with this patch
Signed-off-by: Leah Rowe <leah@libreboot.org>
with this new function, i can turn commands on and
off by virtue of config. for now, behaviour is
unchanged, but this new design means i will be
able to disable certain commands in child processes
Signed-off-by: Leah Rowe <leah@libreboot.org>
This reverts commit 1b65e8914c.
NOTE: cleanup of xbmk_git_init was retained
this "cleanup" actually made the code harder to read.
Signed-off-by: Leah Rowe <leah@libreboot.org>
this was an intense audit of nvmutil that somehow evolved
into writing a new, hardened implementation of mktemp.
it all works, a few memory bugs to solve on bsd, but i don't
see the point in keeping it. mktemp is fine, and nvmutil
already works.
lbutils implemented atomic writes and integrity checking,
in a manner completely overengineered for what it was
actually doing (modifying a few bytes in 8KB GbE files)
just delete it. i'll bring it back if i ever finish the
code. i don't want to leave dead/unfinished code in the tree.
Signed-off-by: Leah Rowe <leah@libreboot.org>
this will make managing updates easier. which will
be soon, because i have a lot of patches.
no idea why i didn't add this before. normally, i just
copy a directory and git-fetch, then cherry-pick as
i please. this will make it trivial to just update
a revision and merge the patches. then i can just
intervene only when necessary, on a given src tree.
Signed-off-by: Leah Rowe <leah@libreboot.org>
it currently boot loops like on some kabylake machines
really buggy codebase. i should just rip off the band-aid
and add edk2 already.
Signed-off-by: Leah Rowe <leah@libreboot.org>
the previous patch was half the battle. now that
annoying person who complained on irc should be
happy.
someone complained about having to enter git
name/email when just doing e.g. ./mk inject
a valid complaint, but their attitude kinda
stunk. oh well. this patch should satisfy them.
Signed-off-by: Leah Rowe <leah@libreboot.org>
Leah Rowe
Leah Rowe
folkmaster
Zoomm Deka