the cbfs function will call cbfstool, which will perform
the same check, and the same error condition would cause
the same exit behaviour in lbmk. the error message would
also provide output that is just as useful for debugging.
Signed-off-by: Leah Rowe <leah@libreboot.org>
There was no error handling, *at all*, on the actual tar
command, due to the lack of set -o pipefail, which we cannot
rely on in sh.
The x_ wrapper can be used in this case, as a mitigation.
Signed-off-by: Leah Rowe <leah@libreboot.org>
We can't do set -o pipefail in POSIX sh, which we're using,
but the build system has x_ which wraps around a command
and executes it, exiting with non-zero status if it does.
This fact enables lbmk to have functionality that is actually
superior to pipefail, since you can more easily control
specifically which parts error.
For example:
foo | bar | foo2 | bar2 | $err "error"
ERROR exits with non-zero status, but foo2, bar and foo
would not exit on error, only bar2 would. In *bash*, which
we avoid, set -o pipefail would make all of them exit on
error, but what if you wanted "bar" to not exit?
With lbmk, you could do, in the above example, and with the
above question asked ("what if you wanted bar not to exit"):
x_ foo | bar | x_ foo2 | bar2 > file | $err "error"
of course, you could also do, if not outputting to "file":
x_ foo | bar | x_ foo2 | x_ bar2
NOTE: in lbmk, $err is a variable containing the name of
a function that does something (whatever you want) and
then exits with non-zero status.
This entire explanation is beyond the scope of simply
providing (and explaining) this fix, but I also wanted to
use this commit as an example of the power of lbmk with
regards to POSIX shell scripting.
Signed-off-by: Leah Rowe <leah@libreboot.org>
x_ cannot be used, where output is redirected to a file;
only the convention piping can be used, for errors.
relying on x_ in these cases will cause unpredictable bugs.
Signed-off-by: Leah Rowe <leah@libreboot.org>
i can't call $err (variable), because it's set
to fail_inject. fix this infinite loop, which
was an oversight in the previous commit.
Signed-off-by: Leah Rowe <leah@libreboot.org>
I was using a complicated method of knowing whether
the current instance was parent or a child, to know
whether the lock file and TMPDIR needed to be purged.
It was quite error-prone too. Instead, I'm now handling
it directly from within the if statement that previously
initialised xbmk_parent=y, forking ./mk from there.
The forked instance would not trigger that if clause
again, since then TMPDIR is created, thus avoiding
recursion.
This is an improvement because it doesn't rely on how
the parent handles exit statuses, and it ensures that
the lock/tmp files are never accidentally deleted.
Even if a given program/script that cbmk runs would
export TMPDIR, it doesn't matter because cbmk doesn't,
so it would be unaffected.
Signed-off-by: Leah Rowe <leah@libreboot.org>
because the top-down function order isn't as reliable
in lib.sh, since this is what first runs, included
in every other script
Signed-off-by: Leah Rowe <leah@libreboot.org>
script/ no longer exists. this means that the
only executable script in cbmk is now mk.
script/trees was never called directly; instead,
we used ./update trees in the past, then just ./mk.
this is part of a larger audit to simplify cbmk,
in preparation for the next Libreboot release.
Signed-off-by: Leah Rowe <leah@libreboot.org>
write to .version and .versiondate, instead
of version and versiondate.
this will hide them to avoid visual clutter while
analysing files within cbmk.
Signed-off-by: Leah Rowe <leah@libreboot.org>
i'm removing all the backward-compatibility in the
build system, so that only the ./mk command is available
Signed-off-by: Leah Rowe <leah@libreboot.org>
Otherwise, the current return prevents set -u -e
after the case/switch block, which is a problem if
set +u +e was done at any point before the return.
Remove the return in the roms) section of the case/switch
block, and make the building of coreboot images part of
an else clause.
Signed-off-by: Leah Rowe <leah@libreboot.org>
this way, the error message will never be incorrect,
which i had to fix in a recent patch.
now, the same string is used for error messages and getopt.
Signed-off-by: Leah Rowe <leah@libreboot.org>
this is not necessary. the fetch mode is still handled,
as before, and no make commands will run in this case.
Signed-off-by: Leah Rowe <leah@libreboot.org>
instead of running pwd all the time, run it once in lib.sh,
and export PWD.
for cbmk-specific use of PWD, use xbmkpwd, which contains
the value of PWD as was set by the pwd utility in lib.sh.
many parts of cbmk rely on pwd, and it *must* be correct.
this change adds basic error handling, since pwd can in
fact return errors in some cases.
Signed-off-by: Leah Rowe <leah@libreboot.org>
it's incorrect for PATH not to be set, but some users
may foolishly blank it out before running cbmk.
prevent such issues, by initialising it.
Signed-off-by: Leah Rowe <leah@libreboot.org>
PWD could be anything, if the user manually exported
it before running cbmk.
always run pwd instead, to get the real string.
Signed-off-by: Leah Rowe <leah@libreboot.org>
several code lines were condensed together, which
make them less readable. make the code more readable
by having separate commands on separate lines.
i previously did this during my manic build system
audits of 2023 and 2024; condensing lines like this
is overly pedantic and serves no real purpose.
Signed-off-by: Leah Rowe <leah@libreboot.org>
Leah Rowe