Here you go - but as I mentioned earlier - extracting the relevant pieces
of the application is next to impossible - and I am not getting the same
errors when running microhttpd's
`make check`.
g++ -m64 -O3 -DNDEBUG -fdata-sections -ffunction-sections -s -flto -fPIE
-fPIC -fstack-protector-strong -D_FORTIFY_SOURCE=2 -std=c++20 -Waddress
-Waggressive-loop-optimizations -Wall -Wcast-align -Wcast-qual
-Wchar-subscripts -Wcomment -Wconversion -Wdisabled-optimization
-Wdouble-promotion -Werror -Wextra -Wfatal-errors -Wfloat-equal
-Wformat-nonliteral -Wformat-security -Wformat-y2k -Wformat=2 -Wimport
-Winit-self -Wno-inline -Winvalid-pch -Wlogical-op -Wmissing-braces
-Wmissing-declarations -Wmissing-field-initializers
-Wmissing-format-attribute -Wmissing-noreturn -Woverlength-strings
-Wparentheses -Wpedantic -Wpointer-arith -Wredundant-decls -Wreturn-type
-Wsequence-point -Wshadow -Wsign-compare -Wsign-conversion
-Wstack-protector -Wstrict-aliasing -Wstrict-aliasing=2 -Wswitch
-Wswitch-default -Wswitch-enum -Wtrigraphs -Wundef -Wuninitialized
-Wunknown-pragmas -Wunreachable-code -Wunsafe-loop-optimizations -Wunused
-Wunused-function -Wunused-label -Wunused-parameter -Wunused-value
-Wunused-variable -Wvariadic-macros -Wvolatile-register-var -Wwrite-strings
-Wno-deprecated-declarations -Wc++11-compat -Wconversion-null
-Wctor-dtor-privacy -Weffc++ -Winvalid-pch -Wold-style-cast
-Woverloaded-virtual -Wredundant-decls -Wsign-promo -Wstrict-null-sentinel
-Wno-useless-cast -Wzero-as-null-pointer-constant -Wstack-protector
-fstack-protector-all -Wno-non-virtual-dtor -Wno-reorder
-Wno-sign-conversion -Wno-float-equal -Wno-old-style-cast
-Wno-zero-as-null-pointer-constant -m64 -Wl,--gc-sections -s -flto -pie -z
noexecstack -z relro -z now -fuse-ld=gold -s main.cpp.o -o bin/main
myapplication_libs.a -lmicrohttpd -lcurl -lgnutls /usr/lib64/libz.so
-lpthread -lgcrypt -lgpg-error -lhogweed -lnettle -lgmp
src/microhttpd/connection.c: In function ‘connection_alloc_memory’:
src/microhttpd/connection.c:281:10: error: ‘need_to_free’ may be used
uninitialized [-Werror=maybe-uninitialized]
281 | if (c->read_buffer_size - c->read_buffer_offset >=
need_to_free)
| ^
compilation terminated due to -Wfatal-errors.
lto1: all warnings being treated as errors
On Thu, 28 Dec 2023 at 06:26, Christian Grothoff <groth...@gnunet.org>
wrote:
> I think it might be helpful if you could share the *exact* compiler
> invocation (make V=1) and error message you are getting...
>
> On 12/27/23 11:38, Sergey Sudnitsyn wrote:
> > Thank you.
> > I don't know what's so special about my setup - I am getting the same
> > error with GCC 12 & 13 - compiling for both x86_64 & aarch64.
> > Compiler flags - nothing unusual - -O3 -flto - but maybe warnings are a
> > bit overzealous (-Wuninitialized -Wstack-protector).
> > I'll try producing a sample that reproduces this - and will use
> > suggested workaround in the meantime.
> >
> > On Tue, 26 Dec 2023 at 17:21, Evgeny Grin <k...@yandex.ru
> > <mailto:k...@yandex.ru>> wrote:
> >
> > Hi Sergey,
> >
> > Are you using some special compiler or linker parameters? Some
> special
> > compiler or linker versions?
> >
> > I regularly test MHD with LTO. I've retested the current git master
> > with
> > LTO and cannot reproduce.
> > No errors, nor even warnings.
> >
> > If you check the code you will find that both mentioned variables are
> > not used before initialisation. (Please provide the code path if I'm
> > wrong).
> >
> > Unless variables are really used as uninitialised (which should be a
> > bug
> > and needs to be fixed), I'd avoid unneeded initialisation to save the
> > code size (and negligible performance).
> >
> > If your compiler/linker complaint is false positive, you probably may
> > workaround it with '-ftrivial-auto-var-init' parameter.
> >
> > Let me know the details, please.
> >
> > --
> > Evgeny
> >
> > On 26/12/2023 16:14, Sergey Sudnitsyn wrote:
> > > When LTO is enabled, linking fails complaining about potentially
> > > uninitialized variables;
> > > It's hard to provide a small sample reproducing this, but I hope
> the
> > > patch can be applied even without it since it's innocent enough
> > (diff
> > > against commit 8bb481f ):
> > >
> > > diff --git a/src/microhttpd/connection.c
> > b/src/microhttpd/connection.c
> > > index f9892a55..e5dd1a45 100644
> > > --- a/src/microhttpd/connection.c
> > > +++ b/src/microhttpd/connection.c
> > > @@ -653,7 +653,7 @@ MHD_connection_alloc_memory_ (struct
> > MHD_Connection
> > > *connection,
> > > {
> > > struct MHD_Connection *const c = connection; /* a short alias
> */
> > > struct MemoryPool *const pool = c->pool; /* a short alias */
> > > - size_t need_to_be_freed; /**< The required amount of additional
> > free
> > > memory */
> > > + size_t need_to_be_freed = 0; /**< The required amount of
> > additional
> > > free memory */
> > > void *res;
> > >
> > > res = MHD_pool_try_alloc (pool, size, &need_to_be_freed);
> > > @@ -4468,7 +4468,7 @@ process_request_body (struct MHD_Connection
> > > *connection)
> > > { /* Need the parse the chunk size line */
> > > /** The number of found digits in the chunk size number
> */
> > > size_t num_dig;
> > > - uint64_t chunk_size;
> > > + uint64_t chunk_size = 0;
> > > bool broken;
> > > bool overflow;
> > >
> >
>
>
