Found a clang bug because of a GCC bug ... Not shocking especially in the area of variadic templates/packs and that the bug in clang is a regression due to a change introduced in C++26 for pack indexing. Which itself introduced an incompatibility between C++26 and before. Just now clang does it always the C++26 way even for C++11.
If I was Biden today alone, I would do the following: * Openly send troops to Ukraine. * Stop all money flow into Isreal and start sending money directly into Palestine * Resign
As a genderqueer Black person with the capacity to give birth living in Texas, I'm finding a lot of you white liberal cishets to be overly dramatic. You just woke up in the same country we had yesterday. Instead of despairing that you don't know what to do, pull your self together and get in where you fit in. Stop depending on others to do your political work for you and become involved beyond telling people who they should vote for every 4 years.
@kevinriggle@Unixbigot I learned that way but instead of directly into BASIC, I turned the code from BASIC to C; I didn't have a basic compiler at the time only a C one.
Building glibc no time at all, running the testsuite hours. Now I was not sure if doing `make check` could use `-j12` or not. For GCC it does the right thing though.
@fclc@phoronix Hmm, I didn't notice they also cut their SVE vector length in half from 512 down to 256 either. Anyways one of the reasons why folks are adding SME is that the SVE streaming vector length (while SME is turned on) can be different from the SVE vector length (when SME is off). But with already 256 in vector length, SME might not be a needed benifit just yet.
Time to do some magic with predicated VN. GCC seems to be able to handle predicated Value Numbering than LLVM too. Though GCC fails in some cases by not looking to see if the conditional was a bitwise and to see if there was more predicates to assert. This is the only second time I have touched SCCVN in GCC (for a missed optimization; the other 2 times were either for an ICE or wrong code). The first missed optimization was back in 2009 dealing with VCE (View Convert Expressions).
@foolishowl@sam It is still suprising how many of socialists tend to shy away and even do the mainstream thing of making fun of folks who use open source software still. What really frustrates me is that many of the same open source tools that are being asked to be used are the basis for many of the closed source products; this whole thing about stealing labor only goes up to a point without pushing for bigger changes.
@whitequark I see the wiki page is only from 2015 while the usage in libstdc++ was basically remove in 2001 (3.0 was released then). This means someone sat down and wrote down what was done over 15 years ago to make sure it was documented correctly on how it worked. Even though it was no longer in use or could be used. From what I read on the glibc list recently is they are looking to see if they can change some of this libio stuff. E.g. https://sourceware.org/pipermail/libc-alpha/2024-September/159942.html
@rini GNU binutils is standalone from GCC. So you are not exactly depending on GCC. It was known to compile at one point with many different compilers (including but not limited to Solaris's CC, HPUX's CC, AIX's xlc). It is still a seperate project from GCC even though they do share the same toplevel configure and build system. I am not sure if that qualifies for what you want though.
@feedable@goldstein So this is a regression too from GCC 5 which didn't have check_flexarrays. I will try to get to a full bug report Tomorrow or Sunday. But check_flexarrays is only the C++ front-end one. The C front-end issue has to be different but maybe a similar issue. I will get one for each then.