| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
autogen needs to be invoked with $(srcdir)/<FOO>-args.def or else it
will not be able to find the input file if GnuTLS is built out of tree,
e.g.
mkdir build
cd build
../configure
make
Also, add missing targets for %-args.h, to avoid this error:
make[2]: Entering directory `/home/user/gnutls/src'
autogen srptool-args.def
autogen psk-args.def
make[2]: *** No rule to make target `ocsptool-args.h', needed by `all'. Stop.
make[2]: Leaving directory `/home/user/gnutls/src'
make[1]: *** [all-recursive] Error 1
For portability's sake we will spell out the rule for each target instead
of using a GNU '%' pattern rule:
https://www.gnu.org/software/make/manual/html_node/Features.html#Features
Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Several problems were found in this area:
1) Currently, if SRC_DEF_* are undefined, autogen will get invoked with
no input file and it will hang forever waiting for content from stdin:
mv -f enums.texi-tmp enums.texi
mkdir enums
../../doc/scripts/split-texi.pl enums enum < enums.texi
echo stamp_enums > stamp_enums
cd ../src/ && autogen -Tagtexi-cmd.tpl && \
rm -f ../doc/invoke-gnutls-cli.texi && \
../doc/scripts/cleanup-autogen.pl <../src/invoke-gnutls-cli.texi >../doc/invoke-gnutls-cli.texi.tmp && \
mv -f ../doc/invoke-gnutls-cli.texi.tmp ../doc/invoke-gnutls-cli.texi && \
rm -f ../src/invoke-gnutls-cli.texi
<HANG>
Since these documents are @include'd by other documents, it is probably
a good idea to make sure the targets are buildable in case they get
listed as prerequisites.
2) SRC_DEF_* used relative paths which are correct for an in-place build,
but incorrect for an out-of-tree build. They should use something like
$(top_srcdir)/src to resolve the ambiguity.
3) cleanup-autogen.pl was also referenced using a relative pathname,
breaking out-of-tree builds.
4) The non-portable "sed -i" flag was used.
Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
|
| |
|
|
|
|
|
| |
Without gperf, priority-options.h does not get built and this results
in a compile error.
Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
|
| |
|
|
|
|
|
| |
This pulls in upstream commit cb3c90598 (stdint, read-file: fix missing
SIZE_MAX on Android).
Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
| |
Other utilities generate invoke-%.texi from %-args.def, but currently
invoke-psktool.texi is generated from psk-args.def. If we make psktool
conform to the same convention as the other utilities, we can use a
generic pattern to handle all of them the same way.
Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
enums.texi is a generated file so we should not look for it in $(srcdir).
When we do, chaos ensues:
mv -f enums.texi-tmp enums.texi
mkdir enums
../../doc/scripts/split-texi.pl enums enum < ../../doc/enums.texi
/bin/bash: ../../doc/enums.texi: No such file or directory
make[4]: *** [stamp_enums] Error 1
make[4]: Leaving directory `/home/user/gnutls/build/doc'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/home/user/gnutls/build/doc'
make[2]: *** [all] Error 2
make[2]: Leaving directory `/home/user/gnutls/build/doc'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/user/gnutls/build'
make: *** [all] Error 2
Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
|
| |
|
|
| |
Chinchani.
|
| | |
|
| |
|
|
|
|
|
|
| |
in cdk_strlist_add, it would check the strlen of the
'string' parameter before it checked if the parameter was
null.
Signed-off-by Jason Spafford nullprogrammer@gmail.com
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
After commit 6addbc3, specifying --enable-local-libopts unconditionally
replaces the autogen-erated files with their distributed copies, and
substitutes AUTOGEN to false.
The assumption here is that if --enable-local-libopts is not specified,
autogen cannot be installed, and that the distributed copies necessarily
exist. Neither assumption is always correct. e.g. someone building a
32-bit copy of GnuTLS from git with a copy of autogen on their system
will have a 64-bit copy of libopts, and a working /usr/bin/autogen,
but not a 32-bit libopts. Since building autogen depends on Guile,
this is a rather heavyweight pile of gear to require. (You can force
a successful build in this case, but it requires providing
AUTOGEN=/usr/bin/autogen to make(1), which is distinctly inelegant.)
So fix things so that if any of the distributed copies do not exist, we
do not substitute AUTOGEN, so as to let any copy of autogen that configure
found on the system do its job if necessary, while not forcing the user
to link against the copy of libopts which came with that autogen.
Signed-off-by: Nikos Mavrogiannopoulos <nmav@gnutls.org>
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
This extension did not get accepted by IETF so it is
now being removed. The gnutls_range API is kept in case
length hiding is implemented in a different way at some
point.
|
| | |
|
| |
|
|
| |
This broke backwards compatibility. Reported by Andreas Metzler.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
gnutls_x509_trust_list_verify_named_crt().
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
