Because we do not define __unused, as on GNU systems Linux and glibc
providing conflicting symbols, declarations for imported functions that
use that attribute cause build failures when used.
Ref: https://bugs.debian.org/1083196Fixes: #34
This make it possible to declare function arguments in headers as
potentially unused, so that they do not emit warnings. We will use this
instead of the BSD __unused, which we cannot currently enable on GNU
systems due to Linux and glibc having conflicting symbols.
With linker sections GC enabled, we get a test failure in `nlist.c`:
```
nlist: nlist.c:72: main: Assertion `rc == 0' failed.
```
This turns out to be because several sections used by the test can be discarded:
```
ld: removing unused section '.text.func_pub' in file 'nlist.o'
ld: removing unused section '.bss.data_pub_uninit' in file 'nlist.o'
ld: removing unused section '.data.data_pub_init' in file 'nlist.o'
```
Reproduced with `CFLAGS="-Og -fdata-sections -ffunction-sections"` and
`LDFLAGS="-Wl,-O1 -Wl,--as-needed -Wl,--gc-sections -Wl,-z,start-stop-gc"`.
Additionally, `LDFLAGS="... -Wl,--print-gc-sections"` can help with diagnosing
which sections get removed.
We already mark these symbols as `used`, but we need `retain` [0] for them
to survive linker GC too.
[0] https://releases.llvm.org/18.1.0/tools/lld/docs/ELF/start-stop-gc.html#annotate-c-identifier-name-sectionsCloses: !29
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This is an internal implementation detail from AC_SYS_LARGEFILE, which
happened to change with autoconf 2.72. Instead compute our own size
for off_t.
Closes: #28
These macros are available in several systems, and we should not install
the man pages for them, otherwise we might end up shadowing the system
man pages if present.
Add the reasoning for why we use -isystem instead of -I, to avoid
confusing unsuspecting readers of the file.
Prompted-by: Alejandro Colomar <alx@kernel.org>
Both modules expose their functions in err.h, so we need to install it
whenever any of them is being built.
Reported-by: Alejandro Colomar <alx@kernel.org>
Handle the three potential system scenarios:
- system time_t is time64
- system time_t is time32 and supports time64
- system time_t is time32 and does not support time64
Add the explicit time32 and time64 functions when necessary and map
them accordingly for each of these cases.
These functions are used by code in the library, even though these
functions started as GNU extensions, they are present in all BSDs,
so we expose them as part of our interface on AIX.
Explicitly select what to include as part of the target ABI, instead of
letting autoconfiguration potentially break ABI if the system grows
functionality provided by the library.
Make almost all the library selectable per target. Do not install manual
pages for interfaces not included in the library. Control inclusion of
symbols in map file via pre-processor macros, and move the comments
describing the ABI selection to configure.ac.
For now the header files are included as is and filtered through
pre-processor conditionals. Eventually they might get switched to be
autogenerated at build time.
Add a new LIBBSD_SELECT_ABI m4 macro, and use it to setup the ABI
interfaces to expose and the various variables and conditionals
to be used by the build system.
Switch to set the initial values to unknown and then set every one
of the selections explicitly by supported target.
Update comments for rationale for things to DROP, or for why some
selections are enabled.
This was accidentally included in the commit, but should have only been
used during development.
Fixes: commit de124dcafac678351366b0572938398ea7ae93e4
Changelog: silent
The digest function checks where unconditionally requiring the functions
to exist or they would error out. But these functions are not required
on all systems, they depend on the ABI to be exposed.
Merge the existing host_os block for the OS detection with the ABI
selection one, as these are related. This way we will be able to make
some of the latter checks conditional on the selected ABI.
On most systems the err family of functions is already present, but are
missing the errc family of functions, which are also present on some
other systems. Splitting them into separate files will make it easer to
conditionally include one or the other.
These contain the fixes to the error handling logic.
In NetBSD the manual page for strtou.3 is generated from the strtoi.3
manual page applying some substitutions, the problem is that the
cross-references are then lost. We will still keep them separate.
Reported-by: Alejandro Colomar <alx@kernel.org>
Some linkers require the map file definitions to contain only symbols
that are present on the linked object, either in the map file or in the
sym file we generate from the map file.
This is preparatory work to be able to conditionally include symbols
in the man and sym files depending on the ABI definitions.
The code is only making the name_from_id function conditional, and
assumes id_from_name are always to be included, so we need to match
the logic for the man page inclusion.
When referring to another manual page and their section number, we need
to use Xr instead of Fn, otherwise the section number is interpreted as
a function argument. For functions provided by libbsd itself we should
be using the 3bsd section instead of 3.
On macOS, closefrom() only sets the close-on-exec flag, so we cannot
check whether all file descriptors were closed, which means that if
on entry our file descriptor table was filled after the 4th file
descriptor, then we might fail the assertions for the flags for odd
file descriptors which we expect to be closed.
This can easily happen when running the test suite in parallel mode
with «make -j8 check» for example.
Closes: #23
The code is only getting the address, but we might be performing an
addressing that is out-of-bounds. Avoid it and use the address form
instead.
Warned-by: cppcheck (objectIndex)
When the system provides implementations for user_from_uid() or
group_from_gid() we are not using these variables, so better not declare
them.
Fixes: commit 21d12b02112097f0c195dceb1892c95b7b957b36
The <sys/*> headers tend to define things that might be used by other
headers, so while they should be self-contained, it is better to simply
include them first.
This was made conditional, but the code part was accidentally left
untouched due to having ported it locally to use __progname, which
caused build failures on the stock repo.
Fixes: commit 046621d7967e7a0f08ae988bcf7e4cd1b6cf204c
When generating the .sym export file from the .map file, we are not
stripping these comments that are part of the same line as the symbol,
which causes ld(1) implementations to error out. Moving them before
the symbols avoids the need to strip them, as we are only keeping
actual symbol lines.
Get the program name from the COMM field from the proc filesystem.
We could use instead the information from the psinfo binary file under
/proc, but that seems to have a shorter string limit.
Although the function is documented as possibly returning NULL if it
cannot find a known source of information, we should still at least
attempt to port it to any supported system, and otherwise explicitly
mark it as not implementable for such systems if that was to be the
case.
Some ld(1) implementations, such as the one on AIX, do not support using
/dev/null as the output filename for the compiled object.
Use an actual filename that we will then clean up.
This makes sure we include it when expected, alongside the man pages,
and the test cases, and do not accidentally break the ABI if the system
starts providing such interface.
This was placed here to make use of the same AS_CASE, but it does not
really fit with the section. Move it to the more appropriate place, and
detangle the AS_CASE.
We had several cases of code needing a strong alias, so we switch those
to use the new macro. This covers systems that support the alias
attribute and others such as macOS where we need to use assembler
directives to add the alias as the attribute is not supported.
The environ variable is supposed to be defined by the code using it, but
on glibc-based systems it will get defined if we request it, by including
<unistd.h> and defining _GNU_SOURCE.
If the system has configured a lower limit (either soft or hard) on the
number of open file descriptors, the test will fail. Make sure to check
whether we have exceeded that limit and adapt the max number of file
descriptors appropriately.
This code was added to cope with Hurd specific behavior, but it is
causing flakiness on containers on some Linux systems. Only enable
it where it is currently needed to try to get stability back on CI
systems.
Closes: #14
This gets rid of the last BSD-4-clause licensed file in the project.
The man page will probably need to be adapted to the current
implementation, but that can be done piecemeal afterwards.
Closes: #7
While attribution is important, it is not relevant when tracking the
copyright holders for the work. And in any case it still stays in the
relevant source file.
This will mean we cannot use sanitizer support on the Hurd, for which
this function was added to fix the test. But the sanitizer suppression
function attribute is not having any effect, so this is better than
nothing.
The versions used in the BSD macros are unknown, so they emit warnings,
extend or reduce them to use the two digit form. Correct the glibc
version when closefrom(3) got introduced.
When the linker uses --no-undefined-version either specified by the user
or as the default behavior (such as with newer clang >= 16 releases),
a missing symbol definition will cause a linker error if that symbol is
listed in the version script.
The __has_builtin operator is more specific and is supported by GCC
and Clang, while __is_identifier() is less specific and only supported
by Clang, so we should prefer the former whenever it is available, and
only fallback to use the latter when the former is missing and the
latter.
AC_SYS_LARGEFILE in configure.ac is setting needed defines to make
64bit off_t on relevant platforms.
Fixes build on musl:
| src/funopen.c:68:28: error: unknown type name 'off64_t'; did you mean 'off_t'?
| funopen_seek(void *cookie, off64_t *offset, int whence)
| ^~~~~~~
| off_t
Closes: !24
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
FreeBSD first declared the function in unistd.h in 2000:
https://cgit.freebsd.org/src/commit/include/unistd.h?id=9feac5c21886
No other #include line was needed to use the function, but the man page
unnecessarily instructed users to include sys/types.h until 2022-11-29:
https://cgit.freebsd.org/src/commit/?id=5b6f0a5012e9bc37dcb32b57b41e2443a46da620
libbsd first declared the function in unistd.h in 2010 with commit
3fed78e5b08f78256e533788b4bcd6502b0949d7 and inherited the bug from
FreeBSD's man page.
No need to do anything for backwards compat.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The check uses printf, so it needs to include <stdio.h> for
compilers which do not support implicit function declarations.
(They were removed from C99.)
Closes: !23
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This function cannot be easily and (more importantly) correctly ported
without cooperation from the libc stdio layer. We already document that
users should be prepared to have the function not available on some
platforms and that they should ideally switch their code to other
more portable and better interfaces.
Instead of making the build fail, and requiring porters to add
exceptions for something that most probably cannot be ported correctly
anyway, simply print a warning and let it build. This will not be a
regression because on those systems libbsd would have never been built
before.
Prompted-by: Jens Finkhaeuser <jens@finkhaeuser.de>
This means we can add a trailing «\» to every element, so that they
can be removed without requiring modification of other lines, and can
be easily sorted.
Replace the old usage of $(nil) which could possibly end up with junk
added if such variable is ever defined, in the environment.
This makes code using it non-portable, and requires the namespeced
headers from libbsd, instead of any generic system headers. It also
requires more code changes than the overlay mode.
On glibc 2.29 reallocarray() was moved to _DEFAULT_SOURCE.
Closes: !20
Based-on-patch-by: Callum Farmer <gmbr3@opensuse.org>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
We test once whether __GLIBC__ is not defined, so we do not need to test
whether it is on the OR branch afterwards. We decouple the glibc version
restriction check from the _*_SOURCE variable, as that contains an
implicit opposite version check.
The former used to be the reference implementation, but it has been
stagnant to the point of not showing much signs of life. Switch to
the currently active and more complete implementation for references.
On the Hurd a small read(3) might end up (indirectly) copying the data
on the stack, which we will end up finding even when we have cleared
the buffer.
To avoid these side effects, we add a new function, that we force not
to be inlined, so that we can reuse the same stack space, that will
blank any possible stack side effects. This should be portable
regardless of stack growing up or down.
Diagnosis-by: Samuel Thibault <sthibault@debian.org>
Ideally we'd recommend getwline(3), but unfortunately even though it
was part of the ISO/IEC TR 24731-2:2010 draft, it did not make it into
C11 and is not widely implemented.
While we are only doing a pointer difference and not dereferencing it,
it's easier and more correct to do the pointer difference before passing
it to reallocarray().
Warned-by: gcc
While using fully uppercase man page titles has been the usual
convention for a very long time, it is rather ugly and something that
some other projects are switching away from.
Fix the following build failure with big endian SH (e.g. sh4aeb):
,---
In file included from nlist.c:44:
nlist.c: In function '__elf_is_okay__':
local-elf.h:223:25: error: 'ELFDATA2LMSB' undeclared (first use in this function); did you mean 'ELFDATA2LSB'?
223 | #define ELF_TARG_DATA ELFDATA2LMSB
| ^~~~~~~~~~~~
`---
Fixes: http://autobuild.buildroot.org/results/2980fb79c208454195d77383f1ece9afbd7f981bCloses: !19
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The arc4random() support for OpenBSD does not build. The intention was
to include any portability code so that the library could be easily
ported to such systems, but it makes little sense to build it on OpenBSD
where most of the functionality will be already present, or the software
would have been ported anyway.
[guillem@hadrons.org: Reword commit message to add rationale. ]
Closes: !15
Signed-off-by: Guillem Jover <guillem@hadrons.org>
- Remove trailing spaces.
- Declare file-scope functions and variables static.
- Declare functions with a proper prototype.
- Do not mix declarations and code for C90 conformance.
- Do not compare size_t and ssize_t variables.
Streams opened with for example open_memstream(3) will have no associated
file descriptor, and fileno(3) will fail.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This does not work with libcs that do not declare the structure
in a header file, like musl. And gets in the way of supporting
non-fd based streams.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This reverts commit beafad2657c7a57109c28f8bad9cb028c84c7dd5.
This test was already handled later on as part of the BUILD_LIBBSD_CTOR
conditional. Adding it to the unconditional set made the build fail when
the system does not have GNU .init_array section support.
Closes: #9
Analysis-by: Duncan Overbruck <mail@duncano.de>
We should not assume that something will implicitly check for this tool,
as we need it ourselves, and this is an internal implementation detail
of right now libtool.
Fixes: commit f11ab6722367f1cf62704ed3c827b9b68dcb5397
We already search for it in `./configure` so let's respect
the result of that search.
This helps with cross-compilation and any other cases
where one might want to choose a different toolchain.
Closes: !16
Bug: https://bugs.gentoo.org/831863
Signed-off-by: Sam James <sam@gentoo.org>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Mention the involved function instead of stating an implicit "this
function". Mention libbsd when proposing using an alternative libmd
to make the context clear.
When using the recent dlsym() based wrapper, we are not requiring any
symbol from libmd, as we resolve those dynamically at run-time. We were
ending up linking against libmd because in another part of the code we
require (depending on the architecture) the SHA512 functions for the
getentropy() local implementation. But that function might be provided
by the system libc on some systems, which means we end up not linking
against libmd at all.
To solve this we go back to the previous simpler solution of linking
directly, which had the main drawback of then making programs fail to
link when not specifying -lmd (on platforms that need it). And then
switch the .so link point from a symlink to a linker script, so that we
can inject the -lmd library as-needed. This is similar to what glibc is
doing.
Fixes: commit 31f034e3862debda8615a449b1c11c4d6920dcc7
To be able to rework the md5 deprecation logic, we need to detangle when
we depend on libmd due to requiring MD5 functions, which might be
otherwise provided by libc, or when we require SHA functions for the
internal getentropy() implementation.
The nlist() function is limited to handle ELF binaries of the same class
as size as the size_t of the architecture built.
In addition the SIZE_T_MAX macro is BSD specific, and was falling back
to the wrong constant on most 64-bit non-BSD systems.
Warned-by: gcc
Because some of the symbols are not otherwise referenced, GCC would like
to remove them.
Closes: !14
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Global asm statements (like .symver directives) do not work reliably
in gcc with link time optimization. Use the symver attribute introduced
with gcc-10 to set symbol versions instead, if available.
[guillem@hadrons.org:
- Simplify by using __has_attribute fallback from <sys/cdefs.h>.
- Coding style changes. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Detect as many warnings as possible during configure and enable them
if the user did not supply any, so that any such problem can be spotted
and fixed.
Switch from the previous versioned symbol implementation which required
users to also link against the message digest provider explicitly, or
they would fail to find the symbols, to an implementation that loads
the symbols from the linked library providing the functions using
dlsym(), thus preserving backwards compatibility.
Using various variables means we have to keep these in sync in various
places. Just use a single variable that we can use anywhere where this
is needed.
More important if close_range() is going to be used, since casting
negative values to 'unsigned int' might hide the errors.
[guillem@hadrons.org: Minor coding style fix. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Take most of the changes done in sudo, but preserve the existing local
changes and refactoring.
In addition, refactor pstat implementation into closefrom_pstat(), so
that the code is easier to read, and requires no conditional
declarations.
This brings <sys/queue.h> to the most up-to-date version from FreeBSD,
incorporating 18 commits from the past 5 years (2015-02-24 - 2021-01-25):
$ git log --oneline 9090a24aed70..8d55837dc133 sys/sys/queue.h share/man/man3/queue.3
Only minimal changes compared to the FreeBSD version have been applied
(queue.3 -> queue.3bsd, _LIBBSD_ prefix).
[guillem@hadrons.org: Remove reference to kernel mode in man page. ]
Closes: !12
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This splits the implementation responsibilities, and reduces embedded
code copies, which was one of the driving points with this project to
start with, so it's nice to give a good example.
[guillem@hadrons.org:
- Remove .exe extension from default program name.
- Call reallocarray() once by switching to a «do {} while» loop.
- Minor coding style fixes. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Add recallocarray(), introduced in OpenBSD 6.1, and freezero(),
introduced in OpenBSD 6.2. The former is imported as-is from OpenBSD,
while the latter is the non-malloc-internal branch of the same code (and
also the OpenSSH portable variant).
Both of these originated in OpenBSD, but have also been implemented by
IllumOS, cf. https://www.illumos.org/issues/8546
Documentation for these functions is in malloc(3) upstream, the relevant
parts of which were previously imported in reallocarray(3bsd). Update
reallocarray(3bsd) with the changes that were introduced since, and add
the relevant bits for recallocarray() and freezero(), plus aliases.
[guillem@hadrons.org: Update copyright in COPYING. ]
Closes: !10
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The Intel compiler does not define __amd64__ on x86_64 platforms;
instead, like other compilers, it defines __x86_64__ .
Closes: !8
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Even though man-pages project now includes man pages for system data
types, we still include these for any other system that does not have
them, to provide a self-contained project with code and documentation.
Some systems such as Windows or musl-libc based ones do not have these
BSD extensions. In addition libbsd itself is making use of the warnx()
functions, so we better provide these interfaces in case they are
missing.
Do not depend on the system vwarn() and verr() functions to implement
the *c() variants, as the system might actually lack any of the <err.h>
BSD extensions.
This is a Russian 64-bit LE VLIW architecture named Elbrus
(formerly Elbrus 2000).
[guillem@hadrons.org:
- Place the entry in alphabetical order. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Commit e8d340de ("Remove a.out support from nlist()") introduced a copy
of the definition of nlist from a.out.h. However, as well as having
n_name inside n_un, on the various BSDs n_name could also be accessed
as a direct member of nlist, and this is made use of by FreeBSD's
usr.bin/netstat/main.c. Thus we should also add the same enclosing
anonymous union.
[guillem@hadrons.org:
- Add a minimal unit test. ]
Closes: !4
Signed-off-by: Guillem Jover <guillem@hadrons.org>
When doing a string comparison for a symbol name from the string table,
we should make sure we do a bounded comparison, otherwise a non-NUL
terminated string might make the code read out-of-bounds.
Warned-by: coverity
There are a couple of malloc() calls with unbounded size arguments,
coming from the parsed file. We need to make sure the size is not
larger than the file being parsed, otherwise we might end up with
out of memory conditions.
Reported-by: Daniel Hodson <daniel@elttam.com.au>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
We should check for partial reads, and not continue in those cases,
as we are not retrying them, otherwise we might end up operating on
uninitialized data.
Reported-by: Daniel Hodson <daniel@elttam.com.au>
Based-on-patch-by: Daniel Hodson <daniel@elttam.com.au>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The sh_link members should be >= e_shnum, otherwise we might do out of
bounds read accesses on the shdr array.
Reported-by: Daniel Hodson <daniel@elttam.com.au>
Based-on-patch-by: Daniel Hodson <daniel@elttam.com.au>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The e_shnum must not be 0, otherwise we will do a zero sized allocation
and further processing of the executable will lead to out of bounds
read/write accesses. The e_shentsize must be equal to sizeof(Elf_Shdr),
otherwise we will perform out of bounds read accesses on the shdr array.
Reported-by: Daniel Hodson <daniel@elttam.com.au>
Based-on-patch-by: Daniel Hodson <daniel@elttam.com.au>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
groff(1) has changed the internal layout for the .Lb doc strings, but to
preserve backwards compatibility we cannot simply rename them, we need
to create new aliases so that these will work with old and new macros.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Windows doesn't provide <sys/param.h>. Several libbsd sources require it
for MIN(), and these are useful non-system-specific macros anyway.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
These warnings are not helpful for libbsd.
[guillem@hadrons.org:
- Rename WINDOWS conditional to OS_WINDOWS.
- Add a nil terminator to the AM_CPPFLAGS. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Extend the host OS checks to define an OS_WINDOWS automake conditional if
the host is MinGW-like. This will be useful for future Windows-specific
build tweaks.
[guillem@hadrons.org:
- Rename WINDOWS conditional to OS_WINDOWS. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The .symver directive is ELF-specific. On non-ELF platforms, work around
this with __attribute__((__alias__)) for the default symbol, and ignore
the variant versioned symbols.
Based-on-patch-by: Aaron Dierking <aarond@fb.com>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This is a glibc-specific symbol that has no public declaration. But is
being used by the OpenBSD and this implementation as a hack to avoid
having to link against the pthread library. This interface is at least
included in LSB 5.0 [L], and using pthread_atfork() is otherwise
problematic anyway [P].
[L] <https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/baselib---register-atfork.html>
[P] <http://austingroupbugs.net/view.php?id=851>
One problem is that we were using it whenever __GLIBC__ is defined,
which is supposed to be defined only on an actual glibc, but uClibc
defines that macro, but it does not provide the symbol on its noMMU
variant.
We add a new configure check that will try to link a program that uses
that symbol to make sure it is present.
Closes: !2
Reported-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Clang's __GNUC__ and __GNUC_MINOR__ definitions are not reliable and may
not be defined at all when targeting the MSVC ABI. Use feature-checking
macros when possible or check for __clang__.
[guillem@hadrons.org: Update for __ protected keyword change. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
These headers are not available on Windows. <bsd/sys/cdefs.h> ensures
that __has_include() and __has_include_next() are defined.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Windows doesn't provide S_ISVTX. Prefer not defining it rather than
defining it to something invalid.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This fixes a regression caused by 2d7de18. These types are not available
on all systems.
Fixes: commit 2d7de186e9cb19a756c0630ee85cb3f2d29b3484
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The loop only executes while len > 0, and the trinary operator in the
function argument is checking against len >= 1 which will always be
true.
Warned-by: coverity
The code uses an internal helper function to avoid code repetition. But
to get there, the function takes a pointer to a pointer, so that the few
functions that require returning an allocated buffer can get hold of it
this way.
The problem is that the user might pass a NULL pointer and trigger an
internal allocation even if the functions are not expected to do so.
Add a new internal helper for non-allocations, that will assert that
condition, and make any other function that requires this behavior call
this one instead.
Warned-by: coverity
Commit 993828d84ee (Add flopenat() function from FreeBSD) dropped the
fcntl.h header. This breaks the build with musl libc:
flopen.c: In function ‘vflopenat’:
flopen.c:60:14: error: ‘O_CREAT’ undeclared (first use in this function)
if (flags & O_CREAT) {
^~~~~~~
Restore the fcntl.h header include to fix the build.
Fixes: commit 993828d84eed0468c6c15b2818e534e6b134b8e4
Submitted-also-by: parazyd <parazyd@dyne.org>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Run «apt install» in non-interactive mode, and do not install
Recommends. Replace build-essential, which is rather fat, with gcc
and make. Execute autogen instead of autoreconf directly.
The NetBSD implementations have different prototypes to the ones coming
from OpenBSD, which will break builds, and have caused segfaults at
run-time. We provide now both interfaces with different prototypes as
different version nodes allow selecting them at compile-time, defaulting
for now to the OpenBSD one to avoid build-time breakage, while emitting
a compile-time warning. Later on, in 0.10.0, we will be switching the
compile-time default to the NetBSD version.
Ref: http://gnats.netbsd.org/44977
Fixes: https://bugs.debian.org/899282
We cannot refer to a License short-name if that is not on its own
License paragraph. Split several other common License fields into
their own paragraphs to avoid this problem in the future.
Use EINVAL instead of EDOOFUS. Add a missing synopsis for
pidfile_fileno() in the man page. Move the definition of struct pidfh
from libutil.h into pidfile.c following upstream change.
Includes changes to handle the Linux syscall blocking when there is not
enough entropy during boot, by switching it to non-blocking mode and
falling back to the alternative implementations. Man page URL reference
fixes. Build fixes for Mac OS X.
Fixes: https://bugs.debian.org/898088
uClibc defines EM_OR1K instead of EM_OPENRISC for the OpenRISC ELF
e_machine ID. Use EM_OR1K when EM_OPENRISC is not defined.
This fixes the following build failure:
In file included from nlist.c:44:0:
nlist.c: In function ‘__elf_is_okay__’:
local-elf.h:224:23: error: ‘EM_OPENRISC’ undeclared (first use in this function)
#define ELF_TARG_MACH EM_OPENRISC
^
nlist.c:77:26: note: in expansion of macro ‘ELF_TARG_MACH’
if (ehdr->e_machine == ELF_TARG_MACH &&
^
Signed-off-by: Guillem Jover <guillem@hadrons.org>
On non-glibc based systems we cannot unconditionally use the
__GLIBC_PREREQ macro as it gets expanded before evaluation. Instead,
if it is undefined, define it to 0.
We should also always declare these functions on non-glibc based
systems. And on systems with a new enough glibc, which provides these
functions, we should still provide the declarations if _GNU_SOURCE
is *not* defined.
Reported-by: Jörg Krause <joerg.krause@embedded.rocks>
This is a non-portable header, and we cannot expect it to be provided by
the system libc (e.g. musl). We just need and rely on declaration that
we have defined ourselves in our own <bsd/sys/cdefs.h>. So we switch to
only ever assume that.
Fixes: https://bugs.freedesktop.org/105281
On IA64 this is only the case in the ELF binary, but it gets normalized
when loaded at run-time.
Fixes: https://bugs.debian.org/881611
Signed-off-by: Guillem Jover <guillem@hadrons.org>
We mention that these are now superseded by the glibc implementations,
make the headers cope with already declared functions on glibc-based
systems, and document this in the man pages.
The two arrays might not reference contiguous memory, and assuming they
are does break at least now on GNU/Hurd, which contains an unmapped
memory block between the memory used by the two arrays.
Just check that each element is strictly after the previous one, so that
we know there are no unmapped memory blocks inbetween.
At least on IA64 and PowerPC 64-bit ELFv1, the functions are stored in
the .text sections but they are accessed through a function descriptor
stored in a data section, for example for PowerPC 64-bit ELFv1 that
section is called .opd.
We should take this into account when checking the n_type for the
functions we have requested information from nlist().
Rationale-by: James Clarke <jrtc27@jrtc27.com>
We have moved all man pages to section 3bsd, let's do the same for the
page title, and in all the references for uniformity and to avoid
confusions.
Fixes: https://bugs.freedesktop.org/101545
Add a check for _MIPS_SIM inside the __mips__ #elif to detect mips64el
and use ELFCLASS64 in that case. Note that we can't use defined(__mips64)
here because that is also defined when the n32 ABI is in use, which uses
ELFCLASS32.
Fixes: https://bugs.debian.org/865091
Signed-off-by: Guillem Jover <guillem@hadrons.org>
* ppc64el defines both __powerpc__ and __powerpc64__ but since the
__powerpc64__ #elif is below the __powerpc__ one, it will never be hit.
* Both assumed that powerpc* was big-endian.
Fixes: https://bugs.debian.org/865091
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The offset is not page aligned, which makes mmap() return EINVAL on
Linux. Switch to use pread() which handles unaligned offset and non-page
sized reads, and because we are already loading parts of the executable
by read() calls, so there's not much point in using mmap() anyway.
Some libc libraries do not have an <a.out.h> header. And a.out as an
executable format is very much obsolete on pretty much all currently
supported systems, even if they might still support loading such
objects.
Remove the a.out support to increase portability.
Backport new changes from OpenBSD.
[guillem@hadrons.org:
- Update copyright years in COPYING. ]
References: https://svnweb.freebsd.org/base?view=revision&revision=281135
Signed-off-by: Igor Gnatenko <i.gnatenko.brain@gmail.com>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Some systems do not have these types available, and they are simply
convenience aliases. Instead use the expanded versions which are more
portable.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=101192
Because we were assigning to another unused variable, when building the
check with optimizations enabled, which is the default when using gcc
as the compiler, the variable was being discarded. Instead pass it to
printf() so that it cannot do so.
These are required due to the O_* macro usage, but have passed
undetected on glibc-based systems due to implicit inclusions.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
In older glibc versions (< 2.17) clock_gettime() is in librt. Add a
check for this to avoid build breakage for programs/libraries that
use libbsd on such systems.
Based-on-patch-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
We are calculating the size of the array, and need to pass the size of
each element, not the size of a pointer to an element. Although this
happens to be the same in many cases, this is not a portable assumption.
Warned-by: coverity
In the function fgetwln() there's a 4 byte heap overflow.
There is a while loop that has this check to see whether there's still
enough space in the buffer:
if (!fb->len || wused > fb->len) {
If this is true more memory gets allocated. However this test won't be
true if wused == fb->len, but at that point wused already points out
of the buffer. Some lines later there's a write to the buffer:
fb->wbuf[wused++] = wc;
This bug was found with the help of address sanitizer.
Warned-by: ASAN
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=93881
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The test in test/strmode.c can fail to compile depending on the
optimization flags used.
The constants that are used in this file (S_IFREG etc.) come from the
<sys/stat.h> include file. It seems gcc ignores this error if one
compiles with "-O2" (default), but if one uses no optimization it fails.
Add the missing include and it works all the time.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=93880
Signed-off-by: Guillem Jover <guillem@hadrons.org>
When running tests in parallel (e.g. using `make -j4 check`), the header
tests currently fail due to headers-overlay.sh and headers-system.sh
both generating headers-gen.c simultaneously, resulting in garbled
output. Fix this by using separate C files for the tests.
Signed-off-by: Lukas Fleischer <lfleischer@lfos.de>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Reuse the getentropy code for Linux on the Hurd, which has fallbacks
for when the better interfaces are not present. And remove all the code
that is not supported currently on the Hurd. Ideally the Hurd should
get an equivalent interfaces that does not suffer from the same
problems as /dev/urandom.
The Unix hook should work for most Unix-like systems, move glibc
specific code there and a FreeBSd specific comment, and remove the rest.
Also change the code to always fallback to use the generic Unix code.
This should cover GNU/Hurd and GNU/kFreeBSD among others.
Actually use the local private SHA512 header instead of relying on the
OpenSSL one for no good reason. Add definition for expected macro
SHA512_DIGEST_LENGTH.
Rework arc4random_stir() and arc4random_addrandom() code over the new
internal API, and documentation in the man page. Adapt the code to the
local build system.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=85827
Add support for the NIOS2 soft-core CPU provided by Altera.
Signed-off-by: Marek Vasut <marex@denx.de>
Signed-off-by: Guillem Jover <guillem@hadrons.org>
Cc: Ley Foon Tan <lftan@altera.com>
Cc: Thomas Chou <thomas@wytron.com.tw>
Cc: Walter Goossens <waltergoossens@home.nl>
The funtools project ships a man page with the same, name. And although
it mith probably make more sense to rename the man page there, as BSD
systems will certainly not do so, this is the easiest and fastest way
to avoid a file conflict.
The fparseln() function had the NetBSD uppercase macros stubbed out,
so replace them with the actual stdio ones. The fgetln() function was
missing any locking at all.
Although the current implementation in libbsd is probably one of the
safest ones around, it still poses some problems when used with many
file streams. This function has now a replacement, that is both more
standard and portable. Ask users to switch to getline(3) instead.
Some 64-bit platforms (e.g. Windows 64) have a 32-bit long. So, shifting
1UL 32-bits to the left causes an overflow. This replaces the constant
1UL with (size_t)1 so that we get the correct constant size for the
platform.
Import from OpenBSD.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
These two functions accept no arguments. The prototypes should reflect
this. This change lets the compiler warn about certain (admittedly
silly) mistakes.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
In case the support is not available, just stop building the
libbsd-ctor.a library, which is a nice to have thing, but should not
have been a hard requirement from the start. This should allow to
build libbsd on non-glibc based systems using another libc.
This is a wrapper over the glibc fopencookie() function.
We diverge from the FreeBSD, OpenBSD and DragonFlyBSD declarations,
because seekfn() there wrongly uses fpos_t, assuming it's an integral
type, and any code using that on a system where fpos_t is a struct
(such as GNU-based systems or NetBSD) will fail to build. In which case,
as the code has to be modified anyway, we might just as well use the
correct declaration.
The glibc headers use selective inclusions through the __need_NAME
mechanism to avoid circular dependencies.
The problem is that if we are being overlaid, and have been requested
a partial inclusion, when we pass control to the system header, then
we might miss definitions needed by our own header, resulting in build
failures.
Workaround that by catching current partial requests, and skip the
current inclusion.
The automatic initialization cannot be part of the main shared library,
because there is no thread-safe way to change the environ global
variable. This is not a problem if the initializaion happens just at
program load time, but becomes one if the shared library is directly or
indirectly dlopen()ed during the execution of the program, which could
have either kept references to the old environ or could change it in
some other thread. This has been observed for example on systems using
Samba NSS modules.
To avoid any other possible fallout, the constructor is split into a
new static library that needs to be linked explicitly into programs
using setproctitle(). As an additional safety measure the pkg-config
linker flags will mark the program as not allowing to be dlopen()ed
so that we avoid the problem described above.
Reported-by: Jan Alexander Steffens (heftig) <jan.steffens@gmail.com>
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=66679
Because clearenv() or setenv() might free the environ array of pointers,
we should make sure to copy it so that we can access it later on when
doing the deep copy via setenv().
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=65470
The GNU .init_array support is an extension over the standard System V
ABI .init_array support, which passes the main() arguments to the init
function.
This support comes in three parts. First the dynamic linker (from glibc)
needs to support it. Then function pointers need to be placed in the
section, for example by using __attribute__((constructor)), that the
compiler (gcc or clang for example) might place in section .ctors and
the linker (from binutils) will move to .init_array on the output
object, or by placing them directly into .init_array by the compiler
when compiling. If this does not happen and the function pointers end
up in .ctors, then they will not get passed the main() arguments, which
we do really need in this case.
But this relies on recent binutils or gcc having native .init_array
support, and not having it disabled through --disable-initfini-array.
To guarantee we get the correct behaviour, let's just place the function
pointer in the .init_array section directly, so we only require a recent
enough glibc.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=65029
The ChangeLog file is distributed, and cannot be regenerated outside
of the git repository, so do not remove it in DISTCLEANFILES, and move
the generation code into dist-hook, which also avoids unnecessary
computation during normal builds.
The code in getpeereid() is unlikely to compile as ucred_t is an opaque
struct (ucred_t * works but ucred_t does not). Either you need to give
a pointer initialized to NULL and getpeerucred() allocates a new ucred
or you call it with an allocated ucred as in this patch.
Signed-off-by: Guillem Jover <guillem@hadrons.org>
The function is a duplicate of expand_number(), but covering less
prefixes and with a slightly different function signature.
Spotted-by: Peter da Silva <resuna@gmail.com>
Make the 0.5 version the default, so that code wanting the actual
implemented version can get a proper versioned depdendency. For code
linked against the old version, make it available as an alias.
Use local getprogname()/setprogname() instead of reimplementing them
locally. Use clearenv() if available, not just on glibc. Use bool
instead of _Bool. Use paranthesis on sizeof. Fold the SPT_MIN macro
into spt_min(). Make spt_init() static. Avoid unnecessary gotos.
Taken from NetBSD.
[guillem@hadrons.org:
- Import from NetBSD instead of FreeBSD to get a 3-clause BSD license,
instead of a 4-clause one.
- Define compatibility macros.
- Change library from libc to libbsd and header in man page.
- Add copyright information to COPYING.
- Add symbol to map file. ]
Signed-off-by: Guillem Jover <guillem@hadrons.org>
This avoids buffer overwrites during concurrent or intermixed calls to
fgetln() when using more than one different stream (currently 32), which
the original interface supports natively by using an internal buffer
from the FILE structure. Although this workaround is rudimentary, it
should cover most of the theoretically problematic cases.
This centralizes the setting so there's no duplication anymore,
makes sure the user supplied variables are never overridden, and
are only set when using gcc.
Reported-by: Samuli Suominen <ssuominen@gentoo.org>
This makes sure the install-exec-hook under src works as expected even
when no runtimelibdir was specified, otherwise the symlinks end up
pointing to non-existing targets.
Reported-by: Ryan Mullen <rmmullen@gmail.com>
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.