The previous logic that used a COMMON_*_FLAGS intermediate variable
and then re-assigned CMAKE_*_FLAGS before including each subdirectory
was confusing and ugly. This PR is the right way to do it.
This commit is purely refactoring: built binaries unchanged.
By default the flag is enabled whenever libunwind is found on the
system, with the exception of static build on OSX (for which we can't
install the throw hook #932 due to lack of support for --wrap in OSX
ld64 linker).
This is an attempt to fix build with STATIC=ON on OSX (#932):
[ 95%] Linking CXX executable ../../bin/bitmonerod Undefined symbols for
architecture x86_64: "___real___cxa_throw", referenced from:
___wrap___cxa_throw in libcommon.a(stack_trace.cpp.o) ld: symbol(s) not found
for architecture x86_64
This fixes build of tests with STATIC=ON, which failed with:
/tmp/cc8lNtqY.ltrans12.ltrans.o: In function
`boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::thread_resource_error>
>::rethrow() const [clone .lto_priv.41]':
cc8lNtqY.ltrans12.o:(.text+0x4e): undefined reference to `__wrap___cxa_throw'
The hook is implemented in libcommon, which is not linked into some of the test
binaries. An alternative solution is to link all tests against libcommon,
but that seems worse because it introduces a false dependency (also,
I tried that and for some of the test binaries the linker still failed to
pick up the symol from libcommon, strangely.)
It is not clear why libunbound was added to this in the first place,
since it wasn't here before and #915 doesn't seem to introduce any
new dependency on it.
Tested build with STATIC=OFF (with and without libunbound-dev libunbound8
installed) and STATIC=ON, on Ubuntu Trusty, Debian Jessie, and Arch
Linux. For static builds, beware of #926 and #907.
If this hack was introduced to make it build on some other system
(Windows? OS X?), then it will have to be dealt with, but not this way.
The tests currently issue a warning that
"warning: -fassociative-math disabled; other options take precedence"
The associative math optimization is turned on indirectly by -Ofast.
Apparently, the optimization is forced to be disabled, while compiling
test harnesses generated by Google Test framework.
Unfortunately, there is no -Wno-error=* flag to disable this warning
(see gcc --help=warnings).
An alternative to this patch is to disable the optimization explicitly
with -fno-associative-math, but that seems worse.
Another alternative is to not pass -Ofast for tests build, but we
want the tests to be built with exact same optimization flags as
the code being tested, otherwise the value of the tests is diminished.
Another alternative is to remove -Werror from the entire build, but
it's good to include that flag to preclude people leaving warnings.
A note regarding implementation of not passing -Werror for tests:
I considered filtering out -Werror from CMAKE_{C,CXX}_FLAGS but
that seems to be worse because it's surprizing behavior, to those
reading the code that adds -Werror. It is better to add it for
when it is used and not added otherwise. I also considered relying
on order, adding -Werror after inluding 'tests' subdir, but before
including the other subdirs, but that also seems cryptic to the
reader. So, I settled with the current solution, of explicitly
setting CMAKE_{C,CXX}_FLAGS to different values before including the
respective subdir.
Testing done: compared compiler invocation for non-tests source files
using `make VERBOSE=1` with and without this commit: the only difference
is the position of -Werror. So, this commit doesn't change the binary.
otherwise clang build fails with
../cryptonote_core/libcryptonote_core.a(miner.cpp.o): In function
`std::__atomic_base<unsigned long long>::load(std::memory_order) const':
/usr/bin/../lib/gcc/i686-pc-linux-gnu/6.1.1/../../../../include/c++/6.1.1/bits/atomic_base.h:396:
undefined reference to `__atomic_load_8'
This has no effect on the gcc build.
The one strange thing is that test code like
std::atomic<int> x;
int main() { return x; }
compiles and links without errors with clang, without -latomic. This
alone would suggest that this patch is unnecessary, but that is not the
case. It's not clear exactly why, though. The bitmonero code is
including the same header, but it must be doing something more complex
than in this test code snippet that causes the failure at link time
pasted above. In any case, passing -latomic fixes the problem and
seems safe.
.
This does two things:
1. fixes clang build, which otherwise errors with undefined symbol
'dlsym'.
2. simplifies the cmake script, delegating to cmake to figure
out platform-specific flags for linking against the dl library.
Tested that it builds with:
gcc 6.1.1, STATIC=OFF,i686
gcc 6.1.1, STATIC=OFF,armv7h
clang 3.8, STATIC=OFF,i686
clang 3.8, STATIC=OFF,armv7h
gcc 6.1.1, STATIC=ON,i686
clang 3.8, STATIC=ON,i686
Also tested that stack trace is generated fine on exception on:
i686, gcc 6.1.1, STATIC=OFF
(didn't bother testing all the other platforms/configs)
This should fix the build problem on OSX (#871, #901), but
I don't have OSX, so I could only test Clang on Linux.
Tested on Linux (Arch) with clang 3.7 and 3.8 i686 and ARM:
if -stdlib=c++ is passed to clang, then the build errors
out with <string>,<iostrea>,etc. headers not found. Simply
not passing the arg fixes the problem.
**NOTE**: not tested on OSX.
Keep the working directory (and umask) inherited from
the parent. Otherwise, it's impossible to control
the working directory of the daemon (from systemd, for
example).
Furthermoer, bitmonerod attempts to create logging directories and files
*in current working directory*. This fails due to permission denied and
generates a (caught, nonfatal) exception. Below is the strace with this
patch applied (so, no `chdir("/")`), showing successful opens at `log/`
relative path. Without this patch they fail (sorry, didn't save the
trace).
```
28911 getcwd("/.../bitmonero", 128) = 25
28911 stat64("/var/lib/bitmonero/.bitmonero", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
28911 stat64("/etc/bitmonerod.conf", {st_mode=S_IFREG|0644, st_size=244, ...}) = 0
28911 open("/etc/bitmonerod.conf", O_RDONLY|O_LARGEFILE) = 3
28911 open("/var/log/bitmonero/bitmonero.log", O_WRONLY|O_CREAT|O_APPEND|O_LARGEFILE, 0666) = 3
28911 stat64("log", {st_mode=S_IFDIR|0700, st_size=4096, ...}) = 0
28911 stat64("log/dbg", {st_mode=S_IFDIR|0700, st_size=4096, ...}) = 0
28911 open("log/dbg/main.log", O_WRONLY|O_CREAT|O_TRUNC|O_LARGEFILE, 0666) = 4
```
The reasoning of chdir("/") in order to prevent the daemon from holding
a filesystem in busy state is not compelling at all: the choice of
working directory for the daemon is the user's business not the
daemon's.
Shorten the list of warnings that are reported, but
which are forced to NOT generate an error, via -Wno-error.
Unwhitelist these: strict-aliasing, sign-compare, type-limits
For example, ignoring strict-aliasing warning caused
lots of wasted time diagnosing Issue #847.
The noexcept specs were added to make GCC 6.1.1 happy (#846), but this
one was missing (because GCC did not complain about it on Linux, but
does complain on OSX).
We need ARCH, because it needs to be set for ARM7, ARM6 to be
initialized.
Strangely, on different machines (both ARMv7, Arch), ${ARCH}
var is either empty or 'native'. Handle both cases.
The destructors get a noexcept(true) spec by default, but these
destructors in fact throw exceptions. An alternative fix might be to not
throw (most if not all of these throws are non-essential
error-reporting/logging).
The destructors get a noexcept(true) spec by default, but these
destructors in fact throw exceptions. An alternative fix might be to not
throw (most if not all of these throws are non-essential
error-reporting/logging).
redfish