Announcement

Collapse
No announcement yet.

Linus Torvalds On GCC 4.9: Pure & Utter Crap

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Linus Torvalds On GCC 4.9: Pure & Utter Crap

    Phoronix: Linus Torvalds On GCC 4.9: Pure & Utter Crap

    Linus Torvalds' latest tirade is over the GCC 4.9 code compiler...

    http://www.phoronix.com/vr.php?view=MTc1MDQ

  • #2
    This is why I like Linus, not pussyfooting around, he just tells it like it is

    Comment


    • #3
      "Gcc is crap" - Linus, 2006

      Comment


      • #4
        I am using Debian too and gcc-4.9.1 currently, but i did not building my kernels with CONFIG_CC_OPTIMIZE_FOR_SIZE so -Os, so not got this issue -O2 does not trigger an issue .

        It might be gcc isn't the problem , but because Debian recently switching from eglibc to glibc . -Os was good supported in eglibc, but not in glibc .
        Last edited by dungeon; 07-26-2014, 12:51 PM.

        Comment


        • #5
          Originally posted by dungeon View Post
          I am using Debian too and gcc-4.9.1 currently, but i did not building my kernels with CONFIG_CC_OPTIMIZE_FOR_SIZE so -Os, so not got this issue -O2 does not trigger an issue .

          It might be gcc isn't the problem , but because Debian recently switching from eglibc to glibc . -Os was good supported in eglibc, but not in glibc .
          No, it's a gcc bug introduced in 4.5.0
          The bad compiler versions are 4.5.0 (when debug_insn came in) to 4.8.3 and 4.9.0 and 4.9.1.
          --https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61904

          Comment


          • #6
            So it is fixed

            https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61801

            Not sure why Michael Danzer triggers it: Debian has 4.9.1 up to the SVN 20140724 (r213031) that fix is included there
            Last edited by dungeon; 07-26-2014, 01:15 PM.

            Comment


            • #7
              From my point of view gcc is getting worse with every single major version (meaning the "x" in 4.x). Since 4.6 (which was quite decent IIRC), I have been facing lots of miscompilations caused by gcc emiting wrong code. With gcc 4.9 it even stopped accepting a valid C++ code from one of my projects. On the bright side after pointing out the commit which broke the compilation it was fixed quite promptly.

              Comment


              • #8
                Originally posted by dungeon View Post
                So it is fixed

                https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61801

                Not sure why Michael Danzer triggers it: Debian has 4.9.1 up to the SVN 20140724 (r213031) that fix is included there
                It is now fixed in the 4.8.3-6 as well which has SVN 20140725 (r213051), so not an issue anymore .

                So it is usual stuff in development *shit happens* . It is only funny because Linus gets angry 8 days after fix is applied .
                Last edited by dungeon; 07-26-2014, 02:07 PM.

                Comment


                • #9
                  Linus is turning into an idiot

                  It is open source and when he does not like it can he choose not to use it, or to try and fix it. It is the same for everyone and for all open source projects. Was this not the point from the start?

                  He then does not need gcc-4.9 when gcc-4.8 did the job for him just fine. Not to mention his dislike for LTO, which makes his attempt at switching to 4.9 look incompetent, because 4.9 is all about bringing stability to this new feature. So what exactly is it he wants with 4.9 if it is not just for getting a free update?

                  So now there is a bug in gcc-4.9, and it makes some of the output look silly. So what? The GNU C Compiler Collection is a huge and complex project. Surely it is understandable to every programmer that a bug can trigger all sorts of strange and unexpected behaviours. Why break into a rage over it?

                  Linus needs to take a step back. His behaviour is irrational. He is being rude and emo-rages over other people's code like a 14-year old boy. Linus is showing more often signs of a tyrant and acts like the Bill Gates of the past in fact. Did we not learn anything from this?

                  I hope to see less of his current behaviour on the news, because it is only terrible for the whole community when their leaders turn openly into idiots. From my experience does it not lead to better code, but causes trouble in the community. It is often a sign for a coming change in leadership, because nobody really wants to contribute to a project run by an idiot for long.

                  Comment


                  • #10
                    Originally posted by sdack View Post

                    Linus needs to take a step back. His behaviour is irrational. He is being rude and emo-rages over other people's code like a 14-year old boy. Linus is showing more often signs of a tyrant and acts like the Bill Gates of the past in fact. Did we not learn anything from this?
                    Torvalds and Gates act and behave nothing alike.

                    I hope to see less of his current behaviour on the news, because it is only terrible for the whole community when their leaders turn openly into idiots. From my experience does it not lead to better code, but causes trouble in the community. It is often a sign for a coming change in leadership, because nobody really wants to contribute to a project run by an idiot for long.
                    Linux and OpenBSD have been proving you wrong for over 20 years now.

                    Comment


                    • #11
                      @sdack
                      I understand what you're saying and I think you make some good points, but I think what you're forgetting is, in Linus' eyes, gcc 4.9 was released without even being properly tested. If its showing errors from things that don't really matter, there could be MUCH worse consequences. Linus is on a timed schedule and his job depends on the reliability and consistency of the software he uses. While I think this was a serious over-reaction, he's not wrong.

                      Comment


                      • #12
                        I think Linus lacked a bit of sympathy here or perhaps he's used to haters calling his software garbage. I'm sure he doesn't like it when they do call his software garbage.
                        I can understand when reacting with discomfort and frustration. This is more along the lines of being blatantly offensive though just to piss people off. Seems like a marketing technique or something.

                        Comment


                        • #13
                          Nice Goonies reference.

                          Comment


                          • #14
                            Originally posted by sdack View Post
                            He then does not need gcc-4.9 when gcc-4.8 did the job for him just fine. Not to mention his dislike for LTO, which makes his attempt at switching to 4.9 look incompetent, because 4.9 is all about bringing stability to this new feature. So what exactly is it he wants with 4.9 if it is not just for getting a free update?
                            Linus isn't even using 4.9, he's on 4.8.3. You apparently didn't actually read what he said, because he's just analyzing the generated code someone else reported was causing a kernel crash.

                            So now there is a bug in gcc-4.9, and it makes some of the output look silly.
                            The underlying gcc bug actually exists back to 4.5, it just happened to only trigger on this code in 4.9.

                            Linus needs to take a step back. His behaviour is irrational. He is being rude and emo-rages over other people's code like a 14-year old boy.
                            He wrote 1 email with a few sentences talking about how the code generation here sucks. If you look at the rest of the list and the bug reports, he also wrote about a dozen more where he was quite calm and helpful in trying to fix the problem.

                            This is another case of the media latching on to a sentence or two, taking them out of context, and then convincing people there is something noteworthy about it. If you've never said something sucks before in an email - well, i don't believe you.

                            Comment


                            • #15
                              Originally posted by Rallos Zek View Post
                              Linux and OpenBSD have been proving you wrong for over 20 years now.
                              In so nice company Linus should threat the GCC project with mail bomb the gcc mailing list!

                              Comment

                              Working...
                              X