| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| | |
The runtest() Perl function runs UN*X commands, so it doesn't work on
Windows.
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| | |
If you're not part of the solution, you're part of the problem. To do a
test, just build with check.vcxproj; the precise way to sacrifice a goat
over msbuild to get it to do the equivalent of "{make,ninja} check" when
building with tcpdump.sln is not exactly obvious.
|
| | | |
|
| | |
| |
| |
| |
| |
| | |
That's not the case on Windows - it doesn't support #!. Look for the
perl interpreter and, if we find it, add a rule that runs tests/TESTrun
with the interpreter.
|
| | |
| |
| |
| |
| | |
Most of the time this probably won't be an issue on UN*Xes, but it'll
probably be an issue on Windows.
|
| | |
| |
| |
| |
| | |
It's all Perl; no UN*X-style shell or commands should be needed any
more, and we should no longer depend on there being a Makefile.
|
| | |
| |
| |
| |
| | |
This test seems to work even with "x87" floating point; perhaps it fails
on, e.g., SPARC with some compiler.
|
| | |
| |
| |
| |
| |
| | |
"Normal" is what you'll get with most versions of most instruction sets;
"x87" is what you'll get with code compiled for 32-bit x86 without SSE,
using the x87 instructions, at least with some compilers.
|
| | |
| |
| |
| |
| |
| | |
Hopefully, that will make sure we don't optimize away anything that
will, for example, cause us not to do things differently on 32-bit x86
using the x87 instructions.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
It's All Very Complicated, so mirror what print-lmp.c does - just do a
calculation based on a particular input value and print the result using
the same format print-lmp.c does, and have tests/TESTrun see what that
result is.
Just do that inside tcpdump.c, so we don't need the fptype stuff.
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
Add a --fp-type flag to tcpdump, which causes it to do a floating-point
operation and, based on the result of the operation, prints out
"FPTYPE{n}", where {n} is a number indicating the result.
Have tests/TESTrun run "./tcpdump --fp-type" and set a HAVE_ key based
on that. Run some tests only for FPTYPE1.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
I can't seem to make those issues pop up on my (x86-64) machine, and, if
they do show up on any platforms, the best hack to handle it is probably
to have a special tcpdump flag to force it to do a floating-point
calculation and see what result is generated, and report the result, so
we know what it'll do with the numbers in the test files, and have
tests/TESTrun run tcpdump with that flag.
If those tests *do* fail, we'll know what calculations to do.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
If we're skipping a test because something *is* set, say "XXX set".
If we're skipping a test because something *isn't* set, say "XXX not
set".
(Don't say the same thing in both cases.)
|
| | |
| |
| |
| | |
While at it, build libpcap silently and use parallel build for tcpdump.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
Remove a stray .md file and restore some comments from the removed .sh
files so future folk can see where the conditional tests come from, not
just the mere fact of their existence. Remove the unused fragile shebang
from testfuncs.pm.
[skip ci]
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
1) We don't need libcrypto to check for the header files, and if we say
we do, we get a lot of noise from CMake about policy CMP0075.
2) We need libcrypto, with the full path specified, for library tests;
just linking with -lcrypto might not find it, or might find another
version (thanks, Apple!).
|
| | | |
|
| | |
| |
| |
| | |
This reverts commit 1e842e24abfadaa0cadbd82ea89fe3170ddb3193.
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
Once we've gotten rid of those scripts, this should no longer be
necessary.
Fixes the problem I mentioned in my comment in GitHub issue #820.
Get rid of extra blank lines at the end while we're at it.
|
| | |
| |
| |
| |
| |
| | |
Create the tests/.passed and tests/.failed files used by test shell
scripts; once we've gotten rid of those scripts, this should no longer
be necessary.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently the return value of link-layer dissectors is supposed to be
the length of the link-layer header, so that it can be skipped for -x
and -X.
If a link-layer dissector or a called function throws an exception, it
returns no value, so that length isn't available.
The goal is to change all the link-layer dissectors to be void functions
and dissectors should update a new field of the netdissect_options
structure "link-layer header length" rather than returning it as a value.
In this transition process, the link-layer dissectors will be moved, when
updated, from the uint_printers[] array (named before printers[]) to the
void_printers[] array.
In this transition process, a new field of the netdissect_options
structure, ndo_void_printer (TRUE/FALSE), set in the updated function
lookup_printer(), will permit to choose between the old and new way to
update the link-layer header length.
|
| | | |
|
| | |
| |
| |
| |
| | |
This first working revision runs a straightforward default compilation of
libpcap and tcpdump on three versions of FreeBSD.
|
| | |
| |
| |
| |
| |
| | |
Some script interpreters do not have the same pathname in all operating
systems where they work (e.g. Perl is /usr/local/bin/perl on FreeBSD).
Apply the classic solution to this classic problem.
|
| | | |
|
| |\ \ |
|
| | | | |
|
| |/ /
| |
| |
| |
| | |
added cpanminus and YAML module for OSX
install local::lib to get local path on OSX for YAML module
|
| | | |
|
| | | |
|
| | |
| |
| |
| | |
situations
|
| | | |
|
| | | |
|
