<div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr">On Tue, Jul 3, 2018 at 5:24 PM Marco Castelluccio <<a href="mailto:mcastelluccio@mozilla.com">mcastelluccio@mozilla.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div text="#000000" bgcolor="#FFFFFF">
    <p>Il 04/07/2018 00:52, Xinliang David Li ha scritto:<br>
    </p>
    <blockquote type="cite">
      <div dir="ltr"><br>
        <br>
        <div class="gmail_quote">
          <div dir="ltr">On Tue, Jul 3, 2018 at 4:29 PM Marco
            Castelluccio via Phabricator <<a href="mailto:reviews@reviews.llvm.org" target="_blank">reviews@reviews.llvm.org</a>>
            wrote:<br>
          </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">marco-c
            updated this revision to Diff 154015.<br>
            marco-c added a comment.<br>
            <br>
            I've taken another approach which looks simpler to me.
            Instead of having two lists, I only have a shared list and I
            use a static variable's address as an identifier for a
            dynamic object.<br>
            <br>
            I've also added another test (which would have failed with
            the previous iteration of the patch) which dlopens three
            libraries.<br>
            <br>
            I will file follow-up bugs for things that don't look right
            in the llvm-cov output which are not regressions from this
            patch:<br>
            <br>
            - instrprof-shared-main-gcov-flush_no-writeout.c.gcov: lines
            23, 26 and 30 should not be covered; lines 33 and 35 should
            not be covered (they are considered as uncoverable instead).<br>
            - instrprof-dlopen-func{2,3}.c.gcov: the only line in the
            function should be covered once and not thrice<br>
            <br>
            Regarding removing writeout and only always using
            __gcov_flush, do you see any disadvantage if we do that?<br>
            <br>
          </blockquote>
          <div><br>
          </div>
          <div>writeout function only writes the profile for the current
            dynamic module which is different from the behavior of
            __gcov_flush. The form is expected when shared lib is
            unloaded -- i.e., do not dump other load module's profile
            unless being explicitly told so (to use gcov_flush).  The
            write out is similar to gcc's gcov_exit. In short, we should
            keep it as it is.</div>
        </div>
      </div>
    </blockquote>
    Sorry, what I meant is that we could remove <span class="m_2752616131475934207pl-s">__llvm_gcov_writeout
      and only keep __llvm_gcov_flush (they are defined in
      lib/Transforms/Instrumentation/GCOVProfiling.cpp and passed to
      llvm_gcov_init). We currently have writeout functions and flush
      functions, where the difference is just that the flush function is
      also resetting the counters to 0.<br>
      At exit, instead of calling __llvm_gcov_writeout, we could call
      __llvm_gcov_flush. Since we are at exit, the fact that we reset
      the counters to 0 for the current module doesn't matter (as the
      module is not usable anymore after that).<br>
      At exit we would still call __llvm_gcov_flush only for the module
      that is exiting, at __gcov_flush we would call __llvm_gcov_flush
      for all modules.<br></span></div></blockquote><div><br></div><div><br></div><div>This sounds safe, but this does not belong to this patch.</div><div><br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div text="#000000" bgcolor="#FFFFFF"><span class="m_2752616131475934207pl-s">
    </span><br>
    <br>
    <blockquote type="cite">
      <div dir="ltr">
        <div class="gmail_quote">
          <div> </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            Regarding llvm_gcov_flush that was recently added, should I
            remove it since we can now unhide __gcov_flush?<br>
            <br>
          </blockquote>
          <div><br>
          </div>
          <div>I suggest we keep it -- there is no harm keeping it.
            llvm_gcov_flush can be documented to be always non hidden. 
            There might also be a use case that user want to discover
            the dumper using dlsym, but only want to dump profile for
            that shared library. In that case, a symbol with protected
            visibility is needed which is a wrapper to the write-out
            method.</div>
        </div>
      </div>
    </blockquote>
    Should we make llvm_gcov_flush hidden and __gcov_flush unhidden
    then? __gcov_flush unhidden would be for compatibility with GCC.<br></div></blockquote><div><br></div><div>Looks like Stephane (the only user of the new interface) is ok for you to remove it completely, which is fine by me.</div><div><br></div><div>David </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div text="#000000" bgcolor="#FFFFFF">
    <br>
    - Marco.<br>
    <br>
    <br>
    <br>
    <br>
    <blockquote type="cite">
      <div dir="ltr">
        <div class="gmail_quote">
          <div><br>
          </div>
          <div><br>
          </div>
          <div>David</div>
          <div><br>
          </div>
          <div><br>
          </div>
          <div> </div>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <br>
            <a href="https://reviews.llvm.org/D48538" rel="noreferrer" target="_blank">https://reviews.llvm.org/D48538</a><br>
            <br>
            Files:<br>
              lib/profile/GCDAProfiling.c<br>
              test/profile/Inputs/instrprof-dlopen-dlclose-main.c<br>
              test/profile/Inputs/instrprof-dlopen-dlclose-main.c.gcov<br>
             
            test/profile/Inputs/instrprof-dlopen-dlclose-main_three-libs.c.gcov<br>
              test/profile/Inputs/instrprof-dlopen-func.c.gcov<br>
              test/profile/Inputs/instrprof-dlopen-func2.c.gcov<br>
              test/profile/Inputs/instrprof-dlopen-func3.c<br>
              test/profile/Inputs/instrprof-dlopen-func3.c.gcov<br>
              test/profile/Inputs/instrprof-shared-lib.c.gcov<br>
             
            test/profile/Inputs/instrprof-shared-lib_called-twice.c.gcov<br>
              test/profile/Inputs/instrprof-shared-lib_in-loop.c.gcov<br>
              test/profile/Inputs/instrprof-shared-main-gcov-flush.c<br>
             
            test/profile/Inputs/instrprof-shared-main-gcov-flush_no-writeout.c.gcov<br>
             
test/profile/Inputs/instrprof-shared-main-gcov-flush_shared-call-after.c.gcov<br>
             
test/profile/Inputs/instrprof-shared-main-gcov-flush_shared-call-before-after.c.gcov<br>
             
test/profile/Inputs/instrprof-shared-main-gcov-flush_shared-call-before.c.gcov<br>
              test/profile/Inputs/instrprof-shared-main.c.gcov<br>
              test/profile/instrprof-dlopen-dlclose-gcov.test<br>
              test/profile/instrprof-gcov-two-objects.test<br>
              test/profile/instrprof-shared-gcov-flush.test<br>
            <br>
          </blockquote>
        </div>
      </div>
    </blockquote>
    <br>
  </div>

</blockquote></div></div>