Replace strange $TEST_FUNCTIONS variable/expansion thing with
something more like what we use in nbdkit, a simple tests/functions.sh
script that gets sourced into each test script.
Update the common submodule to get:
commit 8137d47d0e654065391151eb275e3b64f230f6f5
Author: Richard W.M. Jones <rjones@redhat.com>
Date: Thu Feb 13 11:13:55 2025 +0000
mlcustomize, mltools: Replace $TEST_FUNCTIONS
TEST_FUNCTIONS is being removed from libguestfs and guestfs-tools (it
was removed from virt-v2v a while back). Make the same adjustment in
the common submodule.
(and some other commits which are not relevant to libguestfs)
Run this command across the source:
perl -pi.bak -e 's/(20[012][0-9])-20[12][012]/$1-2023/g' `git ls-files`
and remove changes to po{,-docs}/*.po{,t} (these will be regenerated
later when we run 'make dist').
User-Mode Linux was an alternative hypervisor that could run the
appliance, instead of using qemu. It had many limitations including
lack of network, and UML support in Linux has been semi-broken for a
long time. It was also slower than KVM on baremeal in general and had
various corner cases which were much slower including the emulated
serial port which made bulk uploads and downloads painful. Also of
course it lacked qemu-specific features like qcow2 or any
network-backed disk, so many disk images could not be opened this way.
This was never supported in RHEL.
See-also: https://bugzilla.redhat.com/1144197
Before this change the tests ran in about 12m34 and afterwards in
about 6m20, although the real change is more dramatic if you only run
tests from the tests/ subdirectory (as language tests still run serially).
This breaks valgrinding for now, which I intend to fix properly later.
Apply this change across all the shell scripts containing tests.
Additionally this defines the environment variables $abs_srcdir,
$abs_builddir, $top_srcdir, $top_builddir, $abs_top_srcdir and
$abs_top_builddir which can now be used throughout test scripts.
This macro(?) expands to some shell script to source the
tests/test-functions.sh file from its correct location. The intention
is to use this in all tests, but in this commit only the existing
tests which already include test-functions.sh are modified.
Run the following command over the source:
perl -pi.bak -e 's/(20[01][0-9])-2016/$1-2017/g' `git ls-files`
(Thanks Rich for the perl snippet, as used in past years.)
Create a new top level directory called 'utils' and move the
following programs there:
tests/qemu/boot-analysis -> utils/boot-analysis/
tests/qemu/boot-benchmark -> utils/boot-benchmark/
tests/qemu/qemu-boot -> utils/qemu-boot/
tests/qemu/qemu-speed-test -> utils/qemu-speed-test/
Also we only build the boot-analysis program on x86-64 and aarch64,
since it requires custom porting to each architecture.
When (in --verbose mode) we dump the pass data, dump the times in
milliseconds (instead of nanoseconds) so they are consistent with
other output. Also dump the time difference from the previous event.
Useful for detailed debugging/analysis of problems.
Like with the previous commit, this replaces instances of:
if (something_bad) {
fprintf (stderr, "%s: error message\n", guestfs_int_program_name);
exit (EXIT_FAILURE);
}
with:
if (something_bad)
error (EXIT_FAILURE, 0, "error message");
(except in a few cases were errno was incorrectly being ignored, in
which case I have fixed that).
It's slightly more complex than the previous commit because we must be
careful to:
- Remove the program name (since error(3) prints it).
- Remove any trailing \n character from the message.
Candidates for replacement were found using:
pcregrep --buffer-size 10M -M '\bfprintf\b.*\n.*\bexit\b' `git ls-files`
Wherever we had code which did:
if (something_bad) {
perror (...);
exit (EXIT_FAILURE);
}
replace this with use of the error(3) function:
if (something_bad)
error (EXIT_FAILURE, errno, ...);
The error(3) function is supplied by glibc, or by gnulib on platforms
which don't have it, and is much more flexible than perror(3). Since
we already use error(3), there seems to be no downside to mandating it
everywhere.
Note there is one nasty catch with error(3): error (EXIT_SUCCESS, ...)
does *not* exit! This is also the reason why error(3) cannot be
marked as __attribute__((noreturn)).
Because the examples can't use gnulib, I did not change them.
To search for multiline patterns of the above form, pcregrep -M turns
out to be very useful:
pcregrep --buffer-size 10M -M '\bperror\b.*\n.*\bexit\b' `git ls-files`
It was confusing to have a mix of seconds and milliseconds. For all
upstream testing (eg. with SeaBIOS) we are discussing everything in
milliseconds, so use those exclusively.
Add a new test program called 'boot-benchmark'. This is similar to
'boot-analysis' but it simply boots and shuts down the appliance
several times in a row and measures how long it takes, calculating
mean and standard deviation.
These options allow you to control the appliance memory size, number
of vCPUs, and extra kernel options respectively.
Note that using --smp is not usually a good idea. Not only does it
slow down the appliance, but it tends to break the boot analysis
program because it makes runs (more) non-deterministic.
Move it to the test-data directory, giving it a more generic name so it
can be used also for more than just md5; adjust qemu tests accordingly.
This is just code motion, no behaviour change.
GCC has two warnings related to large stack frames. We were already
using the -Wframe-larger-than warning, but this reduces the threshold
from 10000 to 5000 bytes.
However that warning only covers the static part of frames (not
alloca). So this change also enables -Wstack-usage=10000 which covers
both the static and dynamic usage (alloca and variable length arrays).
Multiple changes are made throughout the code to reduce frames to fit
within these new limits.
Note that stack allocation of large strings can be a security issue.
For example, we had code like:
size_t len = strlen (fs->windows_systemroot) + 64;
char software[len];
snprintf (software, len, "%s/system32/config/software",
fs->windows_systemroot);
where fs->windows_systemroot is guest controlled. It's not clear what
the effects might be of allowing the guest to allocate potentially
very large stack frames, but at best it allows the guest to cause
libguestfs to segfault. It turns out we are very lucky that
fs->windows_systemroot cannot be set arbitrarily large (see checks in
is_systemroot).
This commit changes those to large heap allocations instead.
Use (implicitly) parallel tests on new enough automake. The tests run
a bit quicker with this change. On my laptop, I measured 27 mins down
to 18 mins.
Change the ./run function so that ./run --test no longer spools the
test output to a file. That is not necessary when using parallel
tests, since the test harness does the same thing.
Note: This commit removes the $RUN_OUTPUT_FILE functionality. We will
change the CI integration in a future commit so it uses the .trs and
.log files.
The log file was not being closed at the right time, resulting in the
process running out of open files. Also the log filename buffer was
not freed.
This fixes commit 7bbf5cab28.
This option can be used to direct output from each individual
boot test to a different file. Usual usage is:
qemu-boot --log /tmp/output.% -v -x -n 1000
Because of previous automated commits, such as changing 'guestfs___'
-> 'guestfs_int_', several function calls no longer lined up with
their parameters, and some lines were too long.
The bulk of this commit was done using emacs batch mode and the
technique described here:
http://www.cslab.pepperdine.edu/warford/BatchIndentationEmacs.html
The changes suggested by emacs were then reviewed by hand.
libguestfs has used double and triple underscores in identifiers.
These aren't valid for global names in C++.
The first step is to replace all guestfs___* (3 underscores) with
guestfs_int_*. We've used guestfs_int_* elsewhere already as a prefix
for internal identifiers.
This is an entirely mechanical change done using:
git ls-files | xargs perl -pi.bak -e 's/guestfs___/guestfs_int_/g'
Reference: http://stackoverflow.com/a/228797
The gnulib 'error' module uses 'program_name'. On some platforms --
but not Linux / glibc -- it references it as:
extern char *program_name;
This means when you compile libguestfs on non-glibc (eg. Mac OS X)
gnulib requires 'program_name' as an external string reference, which
we don't provide.
This change doesn't define this string reference for gnulib, but it
does change the name of the macro we use to avoid conflicts if we
eventually need to export 'program_name' as a string.
Thanks: Margaret Lewicka
md5sum(1) does not exist everywhere, so wrap it in an own function so
the right implementation can be chosen on each OS. Also, wrapping it
avoid using awk everytime.
All tests run under the ./run binary. For a long time the ./run
binary has set the $PATH environment variable to contain all of the
directories with binaries in them.
Therefore there is no reason to use ../fish/guestfish instead of just
plain guestfish (and the same applies to other built binaries).
This is useful when analysing with `perf' since it allows us to
increase the amount of time in the test versus the amount of time
spent starting the appliance.
