On 10/11/12 8:50 AM, matchew wrote:
what about openjdk builds used for example by ubuntu package manager (default repos)? these builds do not have these debug files (*.diz) at all.

I don't know how Ubuntu builds the OpenJDK repos so I can't
really comment on why those builds don't have debug info files.
Downstream projects can change the default settings and several
of the downstream Linux project have enabled even more debug
info in their builds.


If i am not considering jvm/jdk native code debugging is it safe to just remove them all?

It's your build. If you don't think you need the debug info, then
you can remove it.

Dan



2012/10/11 Daniel D. Daugherty <daniel.daughe...@oracle.com <mailto:daniel.daughe...@oracle.com>>

    On 10/11/12 3:21 AM, Seán Coffey wrote:
    Moving this off discuss mailing list to build-dev.

    Why is ENABLE_FULL_DEBUG_SYMBOLS being set to 1 for many product
    builds now ? It slows down the build and creates increased bundle
    size even though the majority of users do not require this
    functionality.

    The Full Debug Symbols feature will eventually be enabled for all
    OSes for which Oracle generates bits. I think MacOS X is the last
    platform and that work is tracked by:

        7165611 3/3 implement Full Debug Symbols on MacOS X

    The default setting for ENABLE_FULL_DEBUG_SYMBOLS is "1" (enabled)
    because the Full Debug Symbols feature is intentionally enabled in
    all promoted bits for diagnosibility and debuggability. If FDS is
    not enabled when the promoted bits are built, then the debug info
    generated by a rebuild of *exactly the same source* with FDS enabled
    cannot be (reliably) used with the promoted bits.

    You might be saying:

        That's fine for promoted bits, but what about the rest of us?

    The answer there is actually simple. We want our developer private
    builds and automated system builds, e.g., JPRT, to match what Release
    Engineering builds. We don't want RE to be surprised by an integration
    that builds fine when FDS is disabled only to blow up when FDS is
    enabled.

    Similarly, we also don't want SQE/SQA to be surprised by different
    test results with bits built by RE versus bits built with FDS
    disabled,
    e.g., JPRT. Enabling "debug info" in a build changes the compiler
    optimizations and that changes the bits in the binary. Those changes
    in the binary might mask a bug that only shows up after RE has built
    with FDS enabled. Conversely, if you disable FDS in your private
    build,
    you might end up chasing a bug that only reproduces in your private
    build and doesn't reproduce in an FDS enabled build.


    Could we consider flipping the default for
    ENABLE_FULL_DEBUG_SYMBOLS to 0 ? (like I've done for all my build
    scripts?)

    For the reasons I stated above, no we won't change the default for
    ENABLE_FULL_DEBUG_SYMBOLS to "0" (disabled) unless Oracle changes
    the way that promoted bits are built.

    You are welcome to disable the feature in your private builds which
    is why I added the ENABLE_FULL_DEBUG_SYMBOLS flag. However, please
    remember that any testing that you do with those bits won't
    necessarily
    match testing done with the official promoted bits.

    Dan



    regards,
    Sean.

    -------- Original Message --------
    Subject:    Re: How to decrease size of j2sdk_image
    Date:       Thu, 11 Oct 2012 14:57:46 +0800
    From:       Weijun Wang <weijun.w...@oracle.com>
    <mailto:weijun.w...@oracle.com>
    To:         disc...@openjdk.java.net <mailto:disc...@openjdk.java.net>



    You can try set ENABLE_FULL_DEBUG_SYMBOLS to 0.

    -Max

    On 10/11/2012 02:38 PM, matchew wrote:
    >
    >  After successful openJDK7 build (Ubuntu 12.04) i have found that
    >  'j2sdk-image' has 240MB. More than 100MB belongs to one directory:
    >  openjdk7/jre/lib/amd64
    >
    >  Can anyone explain me why it is so big? For example in openJDK7 installed
    >  via package manager this folder has only 18MB
    >
    >  Is there any way to decrease its size?
    >
    >  Thanks in advance
    >  --
    >  Mateusz
    >




Reply via email to