ObjFW  History of configure.ac of merge-in:2d0b2c1

Some distributions pass LDFLAGS and LIBS to configure that are specific
to building a distribution package, which then causes problems when
using objfw-config outside of this environment. file: [b5a3719683] check-in: [0050fce022] user: js, branch: 1.1, size: 62675 [annotate] [blame] [check-ins using] [diff]

Some distributions pass LDFLAGS and LIBS to configure that are specific
to building a distribution package, which then causes problems when
using objfw-config outside of this environment. file: [6e067ef893] check-in: [538ceacc2f] user: js, branch: trunk, size: 63813 [annotate] [blame] [check-ins using] [diff]

It seems Debian defaults to the integrated assembler, which chokes on
calculating the distance between two labels. file: [aa1242302f] check-in: [1277ea5ce5] user: js, branch: 1.1, size: 61099 [annotate] [blame] [check-ins using] [diff]

It seems Debian defaults to the integrated assembler, which chokes on
calculating the distance between two labels. file: [003a5317d9] check-in: [633a7efff5] user: js, branch: trunk, size: 61158 [annotate] [blame] [check-ins using] [diff]

readdir_r can easily lead to stack overflows if the path is longer than
the space reserved for d_name in dirent. While some OSes use a length of
MAXNAMLEN + 1, others use just 1. This could be worked around by always
allocating a buffer of sizeof(struct dirent) + MAXNAMLEN (which would
work only in case MAXNAMLEN is actually the longest readdir_r can
return), but is probably not worth the risk, especially as glibc has
entirely deprecated readdir_r and it is expected to be removed from
POSIX. file: [63f2b5a434] check-in: [c615b62ed8] user: js, branch: trunk, size: 60786 [annotate] [blame] [check-ins using] [diff]

BSD awk cannot handle binary files.
QNX grep cannot handle binary files.
BusyBox tr cannot strip non-printable characters.

This only leaves sed to strip non-printable characters and then using
grep. file: [fa7714f07b] check-in: [22b11fa7e8] user: js, branch: trunk, size: 61189 [annotate] [blame] [check-ins using] [diff]

QNX's grep doesn't support searching in binary files.

BusyBox's tr doesn't support -dc '[:print:]' to remove non-printable
characters from a binary. file: [38175aae93] check-in: [15dc72eb33] user: js, branch: trunk, size: 60914 [annotate] [blame] [check-ins using] [diff]

This makes configure work on QNX. file: [38cd9ece3e] check-in: [979e82e85f] user: js, branch: trunk, size: 60757 [annotate] [blame] [check-ins using] [diff]

This allows using HTTPS with --iri. file: [fc4f8d39e5] check-in: [7162c635dd] user: js, branch: trunk, size: 60570 [annotate] [blame] [check-ins using] [diff]

This only uses endbr32/endbr64 when actually needed and emits the
necessary .note.gnu.property. file: [406fbcd150] check-in: [de552578e7] user: js, branch: 1.0, size: 59899 [annotate] [blame] [check-ins using] [diff]

This only uses endbr32/endbr64 when actually needed and emits the
necessary .note.gnu.property. file: [ff00996552] check-in: [0b441e52d6] user: js, branch: trunk, size: 60522 [annotate] [blame] [check-ins using] [diff]

The `j` isn't necessary, as tail call eliminations can just use x16/x17,
which allow any type. file: [c0c6842cf8] check-in: [21244a7e62] user: js, branch: 1.0, size: 60337 [annotate] [blame] [check-ins using] [diff]

The `j` isn't necessary, as tail call eliminations can just use x16/x17,
which allow any type. file: [43a1e3f56e] check-in: [ddb7ebcf5b] user: js, branch: trunk, size: 60891 [annotate] [blame] [check-ins using] [diff]

Now that we use SSE in some inline assembly, it needs to be at least 16.
However, _Alignof(max_align_t) can be just 8. file: [fbf20e13cc] check-in: [c80075f6e4] user: js, branch: trunk, size: 60198 [annotate] [blame] [check-ins using] [diff]

There is no reason to have this as a shared library (or .framework) as
it's only used for tests. This also means there doesn't need to be a
stable ABI. file: [84c8505c23] check-in: [1cd17a28dd] user: js, branch: objfwtest, size: 60316 [annotate] [blame] [check-ins using] [diff]

So far, it only discovers all tests and runs them and doesn't do much
else. file: [dcf22b00ec] check-in: [10f006be37] user: js, branch: objfwtest, size: 60492 [annotate] [blame] [check-ins using] [diff]

Mbed TLS is never automatically picked as a TLS implementation and
always needs to be manually enabled using --with-tls=mbedtls. file: [17ac08d8ed] check-in: [11b2a17595] user: js, branch: trunk, size: 60540 [annotate] [blame] [check-ins using] [diff]

It's broken in older versions of Clang (e.g. Clang 12). file: [3d6a6f352e] check-in: [d9af65de97] user: js, branch: trunk, size: 59866 [annotate] [blame] [check-ins using] [diff]

OpenBSD uses KAME but doesn't have inet6_getscopeid. file: [9252cb6573] check-in: [703d77b0cb] user: js, branch: 1.0, size: 59556 [annotate] [blame] [check-ins using] [diff]

OpenBSD uses KAME but doesn't have inet6_getscopeid. file: [e4de599d6e] check-in: [841dd6ddef] user: js, branch: trunk, size: 59865 [annotate] [blame] [check-ins using] [diff]

This is more consistent. file: [f54f1e9278] check-in: [acc1972daa] user: js, branch: trunk, size: 59558 [annotate] [blame] [check-ins using] [diff]

They partially work and it would be better to have support included in
ObjFW 1.0. Once Hurd gets fixed, they will magically work then without
requiring a new ObjFW release. file: [14f4b65308] check-in: [9b50999d63] user: js, branch: trunk, size: 58801 [annotate] [blame] [check-ins using] [diff]

Unfortunately, it still needs to remain at -O0, as Amiga GCC is still
buggy. file: [47dc4b89ca] check-in: [0b6d60ee75] user: js, branch: trunk, size: 58892 [annotate] [blame] [check-ins using] [diff]

Release branches should disable it by default while development branches
should enable it by default. file: [bb2f09756e] check-in: [b20268e251] user: js, branch: 1.0, size: 58504 [annotate] [blame] [check-ins using] [diff]

This will be moved to the "amiga-library" branch entirely in order to
not block the 1.0 release, as this is still far from finished. file: [c17993d113] check-in: [f0b0aacf36] user: js, branch: trunk, size: 58542 [annotate] [blame] [check-ins using] [diff]

This removes functionality which will be restored later. file: [b1471bd183] check-in: [43c80a7593] user: js, branch: trunk, size: 59771 [annotate] [blame] [check-ins using] [diff]

This didn't work without problems on all OSes. Since ofatalkcfg is no
longer there, there is no reason to keep this for now. In the future, if
there is a more generic replacement for ofatalkcfg, this can just be
reverted. file: [579103d61e] check-in: [550bd1b7bb] user: js, branch: trunk, size: 59879 [annotate] [blame] [check-ins using] [diff]

Instead, in the future there should be something more generic that is
not limited to AppleTalk. file: [945ea348f4] check-in: [2ef6db8a44] user: js, branch: trunk, size: 60162 [annotate] [blame] [check-ins using] [diff]

It has a tendency to break exceptions, as only TDM-GCC has patches to
make it work to use a .dll while statically linking libgcc.

While in the CLANG64 environment, this is not an issue, the upside of it
is minimal: It only statically links libunwind instead. While libgcc_s
contains some things that aren't always needed, what is provided by
libunwind is always needed. file: [ee213f3b30] check-in: [b87dab1361] user: js, branch: trunk, size: 58297 [annotate] [blame] [check-ins using] [diff]

I only tried fish for a brief while and am not actively using it or
maintaining the completions. file: [af30c6b6de] check-in: [5f8ebdf8b0] user: js, branch: trunk, size: 58312 [annotate] [blame] [check-ins using] [diff]

* The header was wrong due to still thinking about the check I improved
previously.
* It needs -Werror to not work with an implicit declaration.
* It must not use NULL as an argument, because the parameter could be
declared nonnull. file: [d7854aceb5] check-in: [15365e62aa] user: js, branch: trunk, size: 58635 [annotate] [blame] [check-ins using] [diff]

It seems that neither Windows nor macOS support DDP sockets, but only
the layers on top, while neither Linux nor NetBSD support the layers on
top, but only provide DDP.

It seems that DDP is not meant to be used directly, while Linux and
NetBSD only provide the minimum to use netatalk, which is DDP. ObjFW
would need to use the layers on top on Windows and macOS and implement
those layers itself on Linux and NetBSD. This also means that no code
would be shared between Linux / NetBSD and Windows / macOS, as it's
impossible to have one implementation that works for everything. At this
point, the effort required doesn't make sense. file: [9d8d865f65] check-in: [0b4f713964] user: js, branch: trunk, size: 56174 [annotate] [blame] [check-ins using] [diff]

There are still reliability issues with OpenSSL where connections just
hang. file: [9e644529b8] check-in: [b8d47bbbf0] user: js, branch: trunk, size: 57882 [annotate] [blame] [check-ins using] [diff]

This offers more flexibility in the future, e.g. to generate classes
with properties. file: [0ce2b7ad97] check-in: [1ad52a2f2c] user: js, branch: trunk, size: 55924 [annotate] [blame] [check-ins using] [diff]

devkitPro provides a pthreads wrapper that just works. file: [a6c1722c25] check-in: [bef92f92b6] user: js, branch: trunk, size: 55890 [annotate] [blame] [check-ins using] [diff]

This is more standard, albeit -[connectToNetwork:node:port:] sounding a
little bit more like you connect to a network rather than a node. file: [8dd672b002] check-in: [ccc6a77c55] user: js, branch: trunk, size: 55896 [annotate] [blame] [check-ins using] [diff]

This uses openssl3.library instead so that security issues in OpenSSL
can be fixed without recompiling everything. file: [891ebbfd0d] check-in: [9654e02ae0] user: js, branch: trunk, size: 54197 [annotate] [blame] [check-ins using] [diff]

It crashes. file: [0f0c6381c8] check-in: [c255394e60] user: js, branch: trunk, size: 53992 [annotate] [blame] [check-ins using] [diff]

These implementations all depend on additional libraries that are
otherwise not needed for an ObjFW application that does not use TLS,
meaning they would add unnecessary dependencies for a lot of
applications and waste memory. file: [58ceb289a0] check-in: [8f8af474fc] user: js, branch: trunk, size: 52183 [annotate] [blame] [check-ins using] [diff]

This implements OFTLSStream using Apple's Secure Transport. While Secure
Transport is declared deprecated by Apple, Apple so far has failed to
provide a suitable replacement. They recommend Network.framework as a
replacement, however it can neither work on arbitrary sockets, nor can
it do STARTTLS. file: [fc419f2924] check-in: [112eb62f7b] user: js, branch: trunk, size: 51389 [annotate] [blame] [check-ins using] [diff]

Windows 10 supports UNIX sockets since Insider Preview Build 17063, but
only SOCK_STREAM. file: [6fe2736e44] check-in: [d5373ff70f] user: js, branch: unix-sockets, size: 51035 [annotate] [blame] [check-ins using] [diff]

This avoids running them twice (once on make and once on make install)
and potentially running them as root (during make install). file: [fe0c447cd3] check-in: [be8865bcd2] user: js, branch: trunk, size: 50245 [annotate] [blame] [check-ins using] [diff]

The dev suffix can only be dropped once it is released. file: [3903dc9a4a] check-in: [f1d5ac9683] user: js, branch: 1.0, size: 50105 [annotate] [blame] [check-ins using] [diff]

The x86/x86_64 code has been changed back to AT&T syntax, so the
integrated assembler should be fine now. file: [a6c6d58a54] check-in: [c003df4a8d] user: js, branch: trunk, size: 50307 [annotate] [blame] [check-ins using] [diff]

The dash in the version confuses some package managers. file: [cb63e0206a] check-in: [8fb4f1bfce] user: js, branch: trunk, size: 50394 [annotate] [blame] [check-ins using] [diff]

All necessary changes to adjust for the changes made in trunk are
included in the merge commit. file: [80769d4ef2] check-in: [3c88df0ce4] user: js, branch: amiga-library, size: 51067 [annotate] [blame] [check-ins using] [diff]

It has been moved to the "sctp" branch until it is ready. file: [92281c6a81] check-in: [a863ef544f] user: js, branch: trunk, size: 50383 [annotate] [blame] [check-ins using] [diff]

It has been moved to the "invoke-invocation" branch until it is ready. file: [b9427c0b96] check-in: [c71d3d9433] user: js, branch: trunk, size: 50621 [annotate] [blame] [check-ins using] [diff]

HP-UX 11.11 doesn't have a pthread_create symbol in -lpthread, but the
later AC_LINK_IFELSE succeeds. Therefore, check for main in -lpthread,
to check if we have -lpthread at all, and if we do, always link it. file: [6f06ab9bc1] check-in: [d33a1ae00b] user: js, branch: trunk, size: 50692 [annotate] [blame] [check-ins using] [diff]

The linklib generator already completely obsoletes it. file: [f4b0a15726] check-in: [bb25f8b0fe] user: js, branch: trunk, size: 49727 [annotate] [blame] [check-ins using] [diff]

This prevents leaking resources like open files. file: [3bbb9e4b54] check-in: [169832c303] user: js, branch: trunk, size: 51297 [annotate] [blame] [check-ins using] [diff]

Converting a double to an int64_t requires helper functions on 32-bit
PowerPC. file: [7547bf67ed] check-in: [9510e204b4] user: js, branch: amiga-library, size: 52516 [annotate] [blame] [check-ins using] [diff]

ixemul is quite buggy and the native support we have for MorphOS is far
superior. file: [2aa5aa9da1] check-in: [7ab180e9b6] user: js, branch: trunk, size: 51285 [annotate] [blame] [check-ins using] [diff]

Also make sure to not build both when only one is needed. file: [1d482694cd] check-in: [ce924f47f0] user: js, branch: amiga-library, size: 51530 [annotate] [blame] [check-ins using] [diff]

Check for AmigaOS first, as we have a special implementation for it that
uses WaitSelect() instead. This way, we no longer need to match $host_os
to manually disable the pipe() check, which would break if one of the
AmigaOS variants adds pipe() later on. file: [228ea25515] check-in: [d76213374e] user: js, branch: trunk, size: 50237 [annotate] [blame] [check-ins using] [diff]

The else branch of AS_IF was still not working, so instead of wrapping
AC_PROG_OBJC in anything, build a list of potential compilers in a
variable and use that as an argument for AC_PROG_OBJC. file: [a12670e6ad] check-in: [e80c24128f] user: js, branch: trunk, size: 50222 [annotate] [blame] [check-ins using] [diff]

Searching for an Objective-C compiler within a case makes autoconf
generate code that fails in interesting ways. file: [bb5a38c06f] check-in: [1596518f9d] user: js, branch: trunk, size: 50245 [annotate] [blame] [check-ins using] [diff]

clang does not support baserel, which is required for the .library. file: [c2763df7ae] check-in: [7ab6d090a9] user: js, branch: trunk, size: 50173 [annotate] [blame] [check-ins using] [diff]

clang is currently broken when trying to use a GNU-style runtime with
Mach-O, so this should not be included in 1.0. file: [77b8789dbd] check-in: [eb608dc5f1] user: js, branch: 1.0, size: 48579 [annotate] [blame] [check-ins using] [diff]

Otherwise it will pull in autorelease.m / instance.m twice (once in
ObjFW, and once in the runtime). file: [4baa567a0a] check-in: [5eaa7705ef] user: js, branch: trunk, size: 49243 [annotate] [blame] [check-ins using] [diff]

This avoids creating files in directories that are never used by make
and hence leaving files behind after "make distclean". file: [f8e26f1d63] check-in: [7ac3c91102] user: js, branch: trunk, size: 49059 [annotate] [blame] [check-ins using] [diff]

When arc4random() is unavailable, either random() or rand() is used and
both need to be seeded. If of_random() is a macro, it needs to be
(re)seeded every time, as it's unknown whether it has already been
seeded. As it is seeded with gettimeofday() due to the lack of a better
initial seed, this means every call returns the first state for the
current time, which is very predictable. random() and rand() are both
not cryptographic, but this should at least make it a little bit better
now. file: [989ef7fcda] check-in: [c1fe4b2b77] user: js, branch: trunk, size: 49014 [annotate] [blame] [check-ins using] [diff]

This brings the configure script back in sync with the forwarding code,
which moved to vstmdb as fstmfdd is unsupported by Clang's integrated
assembler. file: [aca5876c18] check-in: [168b1e2e3b] user: js, branch: trunk, size: 49163 [annotate] [blame] [check-ins using] [diff]

It breaks building for the iOS simulator. file: [a08bfedbde] check-in: [f4e99b94d2] user: js, branch: trunk, size: 49164 [annotate] [blame] [check-ins using] [diff]

This was happening before the check. file: [9f7ae83b02] check-in: [08f61f1f18] user: js, branch: amiga-library, size: 49004 [annotate] [blame] [check-ins using] [diff]

ofsock is a command line util to use the various socket classes in ObjFW
and connect them to stdio or each other. file: [cec15c8863] check-in: [97cad6d5f2] user: js, branch: trunk, size: 49173 [annotate] [blame] [check-ins using] [diff]

Just checking for _GNU_SOURCE is not enough, because musl still returns
int even with _GNU_SOURCE, despite the musl header explicitly checking
for _GNU_SOURCE. file: [8e938279fe] check-in: [af603291de] user: js, branch: trunk, size: 47571 [annotate] [blame] [check-ins using] [diff]

These can be enabled with --enable-fish-completions. file: [0fbfa8c4b0] check-in: [d13292673e] user: js, branch: trunk, size: 46519 [annotate] [blame] [check-ins using] [diff]

OFAutoreleasePools have long been discouraged in favor of the runtime's
autorelease pools. file: [5446b6deca] check-in: [09be825f88] user: js, branch: trunk, size: 46238 [annotate] [blame] [check-ins using] [diff]

These have been implemented for a while now and seem to work well. file: [960c7cb4a7] check-in: [beb77aaea1] user: js, branch: trunk, size: 46125 [annotate] [blame] [check-ins using] [diff]

amiga-gcc with -g is currently broken. file: [0badaab9a9] check-in: [ceed953f64] user: js, branch: trunk, size: 45742 [annotate] [blame] [check-ins using] [diff]

This is more compatible with the various conventions used on various
systems. file: [7b78df7df0] check-in: [79fe29dbf4] user: js, branch: trunk, size: 45757 [annotate] [blame] [check-ins using] [diff]

sfdc does not support SysV functions and is not part of the MorphOS SDK. file: [fcd4c9ffc0] check-in: [8f9c2b55ce] user: js, branch: trunk, size: 45827 [annotate] [blame] [check-ins using] [diff]

It's buggy on MorphOS 3.10. file: [1d8a6df756] check-in: [2670c4637e] user: js, branch: trunk, size: 45762 [annotate] [blame] [check-ins using] [diff]

Otherwise, it would be necessary to make unwinding work between 68k and
PPC code, and would get even worse with OS4 support.

This also includes using separate .sfd files. file: [07a7d423c3] check-in: [1e70711672] user: js, branch: trunk, size: 45645 [annotate] [blame] [check-ins using] [diff]

Unlike _wstat64(), GetFileAttributesExW() works on drives. file: [c0e8992130] check-in: [2339bc3407] user: js, branch: trunk, size: 45460 [annotate] [blame] [check-ins using] [diff]

This does not fully fix it yet, but at least the socket tests in the
test suite pass on 3DS now. file: [a713ffb1f0] check-in: [ab89c47f42] user: js, branch: trunk, size: 45186 [annotate] [blame] [check-ins using] [diff]

Instead, OFDNSResolver is used for everything now. file: [ca5e9322d2] check-in: [6e9ee122eb] user: js, branch: trunk, size: 44728 [annotate] [blame] [check-ins using] [diff]

libnetwork also exists on Darwin, but is for different things.

While at it, only add the Windows-specific libraries on Windows. file: [0dca5ff784] check-in: [fd6a6cbafd] user: js, branch: trunk, size: 46838 [annotate] [blame] [check-ins using] [diff]

This does *NOT* provide cryptographically secure randomness! file: [4986935f1d] check-in: [a2a06d372a] user: js, branch: trunk, size: 46358 [annotate] [blame] [check-ins using] [diff]