<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    On 21/11/2013 15:54, Volker Simonis wrote:
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">:
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div>But actually I've just realized that it is not need
                  at all, because 'aix_close.c' isn't in the PATH for
                  any other OS than AIX (that could be probably called a
                  feature of the new file layout:) So I'll simply change
                  it to:<br>
                  <pre><span class="">  48 ifeq ($(OPENJDK_TARGET_OS), aix)</span>
<span class=""><span class="">  49   LIBNET_SRC_DIRS += $(JDK_TOPDIR)/src/aix/native/java/net/</span>
  50 endif</span></pre>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    This make sense.<br>
    <br>
    <br>
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div><br>
                  Yes, exactly. I didn't wanted to change too much code.
                  But as the C-Standard states (<a
                    moz-do-not-send="true"
href="http://pubs.opengroup.org/onlinepubs/000095399/functions/malloc.html">http://pubs.opengroup.org/onlinepubs/000095399/functions/malloc.html</a>)
                  "...If size is 0, either a null pointer or a unique
                  pointer that can be successfully passed to free()
                  shall be returned..." it is perfectly legal that
                  malloc/calloc return a NULL pointer if called with a
                  zero argument. This case is currently not handled
                  (i.e. it's handled as an 'out of memory' error) in
                  check_code.c and I agree that this should be fixed via
                  a separate bug.</div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    Yes, let's use a separate bug for this.<br>
    <br>
    <br>
    <br>
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div> </div>
                <blockquote class="gmail_quote" style="margin:0px 0px
                  0px 0.8ex;border-left:1px solid
                  rgb(204,204,204);padding-left:1ex">
                  <div bgcolor="#FFFFFF" text="#000000"> In net_util.c
                    then it's a bit ugly to be calling aix_close_init.
                    Michael/Chris - what you would think about the
                    JNI_OnLoad calling into a platform specific function
                    to do platform specific initialization?<br>
                    <br>
                  </div>
                </blockquote>
                <div><br>
                </div>
                <div>What about renaming 'initLocalAddrTable()' into
                  something like 'platformInit()' and moving the call to
                  'aix_close_init' to a AIX-specific version of
                  'platformInit()' in net_util_md.c?<br>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    I don't have a strong opinion on the name of the function,
    platformInit is fine for now.<br>
    <br>
    <br>
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div>
                </div>
                :<br>
                <div bgcolor="#FFFFFF" text="#000000"><br>
                </div>
                <div><br>
                </div>
                <div>You're right - we could rename it to something like
                  'java_md_unix.c'. But no matter how fancy the name
                  would be, the file would still be in the
                  'src/solaris/bin' subdirectory:( So I think we'd
                  better leave this for a later change when we
                  completely factor  out the Linux/Mac code from the
                  'solaris/' directory.<br>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    I think JDK 9 is a good time to finally tackle this issue (as you
    probably know, this is something that I've brought up on porters-dev
    or build-dev a few times).<br>
    <br>
    <br>
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div>
                </div>
                <div> </div>
                <blockquote class="gmail_quote" style="margin:0px 0px
                  0px 0.8ex;border-left:1px solid
                  rgb(204,204,204);padding-left:1ex">
                  <div bgcolor="#FFFFFF" text="#000000"> Port.java - one
                    suggestion for unregisterImpl is to rename it to
                    preUnregister and change it to protected so that
                    it's more obvious that it supposed to be overridden.<br>
                    <br>
                  </div>
                </blockquote>
                <div><br>
                </div>
                <div>Done. Also changed the comment to JavaDoc style to
                  be more consistent with the other comments in that
                  file.<br>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    Thanks.<br>
    <br>
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div> </div>
                <blockquote class="gmail_quote" style="margin:0px 0px
                  0px 0.8ex;border-left:1px solid
                  rgb(204,204,204);padding-left:1ex">
                  <div bgcolor="#FFFFFF" text="#000000">
                    UnixNativeDispatcher.c - this looks okay (must
                    reduced since the first round), I just wonder if the
                    changes to *_getpwuid and *_getgrgid are really
                    needed as this just impacts the error message. Also
                    might be good to indent the #ifdef to be consistent
                    with the other usages in these functions.<br>
                    <br>
                  </div>
                </blockquote>
                <div><br>
                </div>
                <div>You're right. This change was done before you fixed
                  "7043788: (fs) PosixFileAttributes.owner() or group()
                  throws NPE if owner/group not in passwd/group
                  database" (<a moz-do-not-send="true"
                    href="http://hg.openjdk.java.net/jdk8/jdk8/jdk/rev/f91c799f7bfb">http://hg.openjdk.java.net/jdk8/jdk8/jdk/rev/f91c799f7bfb</a>).
                  After you're fix it was  "automatically" adapted. I've
                  removed the special AIX handling as suggested because
                  I think as well that another error message in the
                  exception won't have any impact.<br>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    Thanks.<br>
    <br>
    <br>
    <blockquote
cite="mid:CA+3eh105jG=-g3czKvamJnnA_1vw5XAWM7puBKa-6caR3KAGVw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div>
                </div>
                :<br>
                <br>
                <div><br>
                </div>
                <div>I'm currently working on it and created "8028537:
                  PPC64: Updated jdk/test scripts to understand the AIX
                  os and environment" for it because I didn't wanted to
                  blow up this change unnecessarily.<br>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    Okay.<br>
    <br>
    So overall I think this patch is in good shape (I have not reviewed
    the AWT/2D/client changes in any detail) and I don't see any
    blocking issues to this going in.<br>
    <br>
    -Alan<br>
    <br>
    <br>
    <br>
  </body>
</html>