Adhemerval Zanella

GNU Tools Weekly News Week 24 (February 8, 2026) Release updates for GNU toolchain: * binutils 2.46 was released * @Gerda.invalid/T/#t" target="_blank" rel="nofollow noopener">inbox.sourceware.org/binutils/87fr7… * gcc 16.0 regression status * P1: 29 * P1+P2+P3 total: 860 General/big GNU toolchain news: * GCC easy issue to tackle of the week: * New feature this week, This is a highlight of one bug report that would be a good issue for someone new to GCC to fix * If others want to sponsor an issue please let me know and I can add that one for the week (gdb, binutils and glibc issues welcome too) * lower mempcpy to memcpy when result is unused * gcc.gnu.org/bugzilla/show_… * Reach out to Andrew Pinski for mentoring on this issue GCC commits: * note most of the commits from now until the release are bug fixes which have not been listed here normally * doc: Move parameter docs to the GCC internals manual * Removal of CONST_CAST and related macros (C++ification cleanup) GCC discussion: * GCC bugzilla stats * 118 new issues filed * 136 issues closed glibc commits: * math: Order signed zeros in f{max,min}{,mag}{f,l,f128} * math: Optimize f{max,min}imum{,_num,_mag,_mag_num}{f,l,f128} * AArch64: Optimize memcpy for Kunpeng 950 processor * AArch64: Add if('fastmath') to math-vector-fortran.h binutils/gdb commits: * PowerPC: Support for Elliptic Curve Cryptography Instructions (RFC02669) * [gold] Note gold and dwp deprecation in NEWS * bpf: add may_goto instruction

The GNU C Library version 2.43 Released. ISO C23 Features. Linux mseal. Additional optimized and correctly rounded mathematical functions. Experimental Clang build support. AArch64 improvements. sourceware.org/pipermail/libc…

@spendergrsec @seanhn Are any of these applicable to upstream glibc or are PAX specific ? I would be interested to take a look if possible.

@seanhn (and pipacs had a ton of patches a decade+ ago for glibc related to all this stuff that I don't believe have been replicated yet)

@seanhn github links in new blog 404 (maybe I'm reading too early :))

@spendergrsec @seanhn No afaik, I finally have added llvm build support on next 2.43 release and plan to check on enable its CFI support.

@seanhn does glibc even have support for fine-grained CFI? Blog mentions CFI in multiple spots, but I think may be referring to different kinds for different code. Code I looked at for the atexit handling doesn't look like it'd work under type-based CFI

@itanium_guy Another problem is afaik DWARF unwinders are inherent slower than a SEH one (need to mmap and parse extra metadata).

@adhemerval Oh yes! For sure! longjmp is multiple times (not percent) slower on Windows than it is Unix based systems. But is is also that many time less used and offers a clean and safe integration with the rest of the code's assumptions (that finalizers always run).

It is 2026 and glibc's longjmp still doesn't execute finalizers when truncating the stack.

@itanium_guy On glibc we can just force libgcc_s.so loading at startup, it would make async-signal in theory (old libgcc still calls malloc on unwind, but it was fixed a long time agora afaik). But you will still the have de extra runtime cost and semantic change.

@adhemerval Of course, in Windows care the unwinder lives in the lowest userland layer (ntdll.dll - even below the crt) so there is no need to dload it.

@itanium_guy The pthread_cancel/pthread_exit C++unwind integration is a glibc extension; afaik other C runtimes do not implement it (musl, BSD).

@itanium_guy I would say not many in the C world where Linux/glibc was initially envisioned. On Windows you have SEH and C extensions where exceptions where designed as system primitive, which is not the case for Linux/glibc.

@itanium_guy And this semantic will be a versioned symbol (to avoid breaking programs that do not expect it), which add more maintanance and complexity. It think you might bring this up on glibc, but I think is unlikely other maintainer will want to change it.

@itanium_guy This would require to dlopen libgcc_s.so, similar to what pthread_cancel do; which would change the function to be not be async-signal-safe. It would also add some overhead on most of users that do not require it.

The entire world is running out of RAM and for once it’s not @googlechrome’s fault. 🙏

GNU Tools Weekly News Week 15 (December 7, 2025) Release updates for GNU toolchain:  * GDB 17.1 release next weekend (Sat-Sun Dec 13-14)   * @adacore.com/T/#u" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gdb-patches/aT…  * glibc 2.43 release planning   * @redhat.com/T/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/libc-alpha/f8f…   * sourceware.org/glibc/wiki/Rel… General/big GNU toolchain news:  *  The GCC Steering Committee has a wiki page detailing their mission  and current requests are listed   ** gcc.gnu.org/wiki/gcc-steer…  * What’s New in Oracle Solaris 11.4 SRU 87   ** GCC 15 included now   ** blogs.oracle.com/solaris/whats-… GCC commits:  * Fixes for aarch64 mingw support (long double and varargs)  * MIPS: add support for mips allegrex cores  * GCC rust: 12 patch sync  * Ada: 7 patch sync  * Sync libsanitizer from upstream  * amdgcn: Implement Managed Memory  * libstdc++: implement P3044R2  * libstdc++: Convertibility of rank == 0 layouts, LWG4272.  * libstdc++: Implement P2655R3  * Add -ffuse-ops-with-volatile-access (enabled for x86 by default)  * OpenMP: C/C++ parser support for dyn_groupprivate   ** Fortran support was earlier GCC discussion:  * injected class name vs. constructor reference (with templates)   ** @connect.xfinity.com/T/#u" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/280584618.…  * "A: a new bug to old plain C"   ** @mail.gmail.com/T/#t" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/CAKBShRQUB…  * Bug in identical code folding?   ** @gjlay.de/T/#t" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/aa96c8d9-e…  * Question about ifcvt optimization   * @mail.gmail.com/T/#u" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/CAPt2h2Ygx… GCC bugzilla stats  * 113 new issues filed  * 79 issues closed glibc commits:  * aarch64: Optimise AdvSIMD log1pf, logf, log2f, log10f, log1p, log, log2, log10  * aarch64: Optimise AdvSIMD acoshf, asinhf, atanhf  * aarch64: Implement AdvSIMD and SVE rsqrt(f) routines glibc discussion:  * Deprecation of s390-32 in glibc   * @linux.ibm.com/T/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/libc-alpha/b1d… binutils/gdb commits:  * gdb/guile: remove support for Guile < 2.2  * Add AMD znver6 processor support

GNU Tools Weekly News Week 14 (November 30, 2025) Release updates for GNU toolchain:  * GDB 17.1 release update -- 2025-11-29   * @adacore.com/T/#u" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gdb-patches/aS… General/big GNU toolchain news:  *  The GCC Steering Committee has a wiki page detailing their mission and current requests are listed    ** gcc.gnu.org/wiki/gcc-steer…  * Reminder: GCC (GNU Toolchain) dev room at FOSDEM 2026, January 31-Februrary 1, Brussels, Belgium - Calls for proposals   ** The deadline is tomorrow December 1, 2025   ** @euclid.schwinge.ddns.net/T/#m7f70eb7b8870e94935003b629d822747ef012ff0" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/87jyzizitk…  * GDB: Christina Schimpe added as a x86-64 and i386 maintainer  * Office Hours for the GNU Toolchain was held Thursday 2025-11-27   * Notes can be found at gcc.gnu.org/wiki/OfficeHou…   * Next office hours is 2025-12-18 @ 1100h EST5EDT * Sourceware HW service migration successfully happened Friday 2025-11-28   * Please report if you notice anything wrong to overseers@sourceware.org  * The Algol 68 front-end for GCC was merged in  * GCC's C++ front-end now defaults to C++20 if not supplied a -std= option GCC commits:  * The Algol 68 front-end  * Change the default C++ dialect to gnu++20  * c: Add _Maxof and _Minof operators  * New g++ option --compile-std-module to compile the std module without any extra invocation of gcc  * RISC-V profiles support was removed from GCC 16 due to issues  * aarch64: Enable parsing of user-provided AArch64 CPU tuning parameters  * aarch64: Better code generation for SVE/Adv. SIMD early break vectorization  * libstdc++: Implement LWG 4366 for std::expected comparisons  * libstdc++: Implement LWG 4370 for std::optional comparisons  * gccrs: 12 commit sync  * ada: 15 commit sync  * RISC-V: Add Andes 23 and 45 series pipeline descriptions  * RISC-V: Add SpacemiT extension xsmtvdot GCC discussion:  * decimal floating point   ** Moved from glibc discussion   * @mail.gmail.com/T/#m2e0d4dd4a417da002fbbc32a50486d1ae6f9d1ca" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/CAH6eHdRpX…  * Refining _DEFAULT_SOURCE and _GNU_SOURCE   ** Also glibc discussion   * @redhat.com/T/#m796bfe7a1c918d55a87995df11747975d71fcba2" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/8926a67d-1… GCC bugzilla stats  * 113 new issues filed  * 97 issues closed glibc commits:  * math: New generic fmaf and fma implementation   ** related to last week's `Remove ldbl-96 fma implementation` commit   ** Also see zatrazz.substack.com/p/the-recent-g…  * math: Sync atanh from CORE-MATH  * Define C23 header version macros  * stdlib: Remove longlong.h (usage and header) glibc discussion:  * See gcc discussion too  * glibc SFrame support for newer versions   * @oldenburg.str.redhat.com/T/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/libc-alpha/lhu…  * Is hppa64-linux-gnu an supported target?   * @mail.gmail.com/T/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/libc-alpha/CAM… binutils/gdb commits:  * Obsolete support for Solaris < 10 gdb discussion:  * What bugs/issues should I work on?   * @mail.gmail.com/T/#u" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gdb/CAB9PnCrXQ…

It seems some forum commenters got confused about the glibc's fma optimization from Phoronix article, let try to explain it better zatrazz.substack.com/p/the-recent-g…

Fangruin raises valid concerns about the SFrame design that creators unfortunately have not acknowledged yet. We reluctantly added this on glibc (it was added near the release) and I wonder if we should wait more.

Not so obvious I must say, but it is still interesting - zatrazz.substack.com/p/not-so-obvio…

GNU Tools Weekly News Update Week 9 (October 26, 2025) General/big GNU toolchain news: * Valgrind 3.26.0 released * Stabs debugging info support was removed from GDB GCC discussion: * Contributing the Algol 68 Front-End to GCC ** @oracle.com/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/87zf9kov10… * Do people find the "Option Summary" in the GCC manual useful? ** @baylibre.com/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/gcc/e9a10fd7-4… GCC commits: * C2y: Implement static assertions in expressions * C2y: Allow unspecified arrays in generic association. * Simplification of vec_perm chains implementing concat-and-extract ** Improves SIMDe code generation for non-x86 targets * Simplification pattern of round semantics ** Improves SPEC 2017's 538.imagick_r by ~3% * C++23: Implement P2674R1 - A trait for implicit lifetime types * Implement bool reduction vectorization ** x86_64 and aarch64 (both SIMD and SVE) * libstdc++: Implement optional from P2988R12 (C++26) * cobol: Implement the XML PARSE statement. * phiopt: Finish up moving minmax_replacement to match-and-simplify ** This was a partial transition that was started 2 years ago GCC bugzilla stats * 93 new issues filed * 89 issues closed glibc commits: * Many fixes for compiling glibc with clang gdb/binutils commits: * GDB: Stabs debugging support was REMOVED binutils discussion: * continued GNU Tools Cauldron SFrame talk notes ** @oracle.com/" target="_blank" rel="nofollow noopener">inbox.sourceware.org/binutils/933bb… ** Also related LLVM discussion about adding SFrame support: discourse.llvm.org/t/rfc-addin

RISC-V BE support ... "Just because you can, doesn’t mean you should”

If you are wondering about latest glibc math work conf.gnu-tools-cauldron.org/opo25/talk/BKR…