Re: [PATCH 4 of 7] cc/gcc: Add support for getting a gcc snapshot

[Bryan Hundven <bryanhundven at gmail dot com>]

On Thu, Dec 9, 2010 at 2:56 PM, Bryan Hundven <bryanhundven@gmail.com> wrote:
> On Thu, Dec 9, 2010 at 2:06 PM, Yann E. MORIN
> <yann.morin.1998@anciens.enib.fr> wrote:
>> Bryan, All,
>>
>> On Tuesday 07 December 2010 10:06:57 Bryan Hundven wrote:
>>> # HG changeset patch
>>> # User Bryan Hundven <bryanhundven@gmail.com>
>>> # Date 1291712194 28800
>>> # Node ID 8f7cc140fe23d204d2d1915fb3f496f4eef9a7c6
>>> # Parent Â73795525c82eaa7f4561f358c70d26ef9d72a446
>>> cc/gcc: Add support for getting a gcc snapshot
>>>
>>> diff -r 73795525c82e -r 8f7cc140fe23 config/cc.in
>>> --- a/config/cc.in  ÂMon Dec 06 22:42:07 2010 -0800
>>> +++ b/config/cc.in  ÂTue Dec 07 00:56:34 2010 -0800
>>> @@ -8,6 +8,9 @@
>>> Âconfig CC_VERSION
>>> Â Â Âstring
>>>
>>> +config CC_SNAPSHOT
>>> + Â Âstring
>>> +
>>> Âsource "config.gen/cc.in"
>>>
>>> Âconfig CC_SUPPORT_CXX
>>> diff -r 73795525c82e -r 8f7cc140fe23 config/cc/gcc.in
>>> --- a/config/cc/gcc.in    ÂMon Dec 06 22:42:07 2010 -0800
>>> +++ b/config/cc/gcc.in    ÂTue Dec 07 00:56:34 2010 -0800
>>> @@ -19,6 +19,30 @@
>>> Â# Don't remove next line
>>> Â# CT_INSERT_VERSION_BELOW
>>>
>>> +config CC_V_4_6_SNAPSHOT
>>> + Â Âbool
>>> + Â Âprompt "4.6 Snapshot (EXPERIMENTAL)"
>>> + Â Âdepends on EXPERIMENTAL
>>> + Â Âselect CC_GCC_4_6_or_later
>>> +
>>> +config CC_V_4_5_SNAPSHOT
>>> + Â Âbool
>>> + Â Âprompt "4.5 Snapshot (EXPERIMENTAL)"
>>> + Â Âdepends on EXPERIMENTAL
>>> + Â Âselect CC_GCC_4_5_or_later
>>> +
>>> +config CC_V_4_4_SNAPSHOT
>>> + Â Âbool
>>> + Â Âprompt "4.4 Snapshot (EXPERIMENTAL)"
>>> + Â Âdepends on EXPERIMENTAL
>>> + Â Âselect CC_GCC_4_4_or_later
>>> +
>>> +config CC_V_4_3_SNAPSHOT
>>> + Â Âbool
>>> + Â Âprompt "4.3 Snapshot (EXPERIMENTAL)"
>>> + Â Âdepends on EXPERIMENTAL
>>> + Â Âselect CC_GCC_4_3_or_later
>>> +
>>
>> I'd rather have the 4.x snapshots be beside the corresponding 4.x.y
>> stable versions, eg:
>> Â4.6-snapshot
>> Â4.5-snapshot
>> Â4.5.1
>> Â4.5.0
>> Â4.4-snapshot
>> Â4.4.5
>> Â4.4.4
>> Â...
>> Â4.3-snapshot
>> Â4.3.5
>> Â...
>
> Sure! Makes sense to me. I think Arnaud's version of the patch did the same.
>
>>> Âconfig CC_V_4_5_1
>>> Â Â Âbool
>>> Â Â Âprompt "4.5.1 (EXPERIMENTAL)"
>>> @@ -154,6 +178,11 @@
>>> Â Â Âselect CC_GCC_USE_MPC
>>> Â Â Âselect CC_GCC_HAS_LTO
>>>
>>> +config CC_GCC_4_6_or_later
>>> + Â Âbool
>>> + Â Âdefault n
>>> + Â Âselect CC_GCC_4_5_or_later
>>> +
>>> Âconfig CC_GCC_HAS_GRAPHITE
>>> Â Â Âbool
>>> Â Â Âdefault n
>>> @@ -188,6 +217,10 @@
>>> Â Â Âstring
>>> Â# Don't remove next line
>>> Â# CT_INSERT_VERSION_STRING_BELOW
>>> + Â Âdefault "4.6-" if CC_V_4_6_SNAPSHOT
>>> + Â Âdefault "4.5-" if CC_V_4_5_SNAPSHOT
>>> + Â Âdefault "4.4-" if CC_V_4_4_SNAPSHOT
>>> + Â Âdefault "4.3-" if CC_V_4_3_SNAPSHOT
>>> Â Â Âdefault "4.5.1" if CC_V_4_5_1
>>> Â Â Âdefault "4.5.0" if CC_V_4_5_0
>>> Â Â Âdefault "4.4.5" if CC_V_4_4_5
>>
>> Ditto.
>
> ack.
>
>>> @@ -217,6 +250,15 @@
>>> Â Â Âdepends on CC_LANG_JAVA
>>> Â Â Âdepends on CC_GCC_4_3_or_later
>>>
>>> +config CC_SNAPSHOT
>>> + Â Âstring
>>> + Â Âprompt "gcc snapshot date"
>>> + Â Âdepends on CC_V_4_6_SNAPSHOT || CC_V_4_5_SNAPSHOT || CC_V_4_4_SNAPSHOT || CC_V_4_3_SNAPSHOT
>>> + Â Âdefault ""
>>> + Â Âhelp
>>> + Â Â Check ftp://gcc.gnu.org/pub/gcc/snapshots/ for the latest snapshot.
>>> + Â Â The string to enter should be just the 'YYYYMMDD' date of the snapshot.
>>> +
>>
>> How long are the snapshots kept on there? It looks like only for about
>> 6-8 months. Which means that starting again with the same .config would
>> make the toolchain no longer buildable after that period of time has
>> elapsed.
>>
>> This should probably not be an issue, as using a toolchain based on a
>> snapshot for production would be highly dubious. But once the option
>> is here, people will start using it, forget about it, and ship their
>> snapshot-toolchain to the outter world. Which means they'll be screwed
>> if they have to fix it later.
>>
>> So, I would like you add a BIG FAT WARNING roughly saying:
>>
>> ÂBIG FAT WARNING ! You are using a snapshot, there is no guarantee that
>> Âthe toolchain will properly work, or even build. You might even not be
>> Âable to reproduce it later, if the upstream snapshot eventually disapears
>> Â(and they have a tendency to disapear after a not-so-long period).
>>
>> ÂSo, it is HIGHLY advised:
>> Â- that you think twice before you keep using a snapshot, and if so:
>> Â- that you save the snapshot tarball by your own means in a safe place.
>>
>> Roughly that. I leave it to you to rephrase it, if you want. :-)
>
> I totally agree.
> Maybe besides having EXPERIMENTAL, we should have SNAPSHOT or DEVELOPER.
> And these options would have the "BIG FAT WARNING". What do you think?
>
>>> Âconfig CC_PKGVERSION
>>> Â Â Âstring
>>> Â Â Âprompt "gcc ID string"
>>> diff -r 73795525c82e -r 8f7cc140fe23 scripts/build/internals.sh
>>> --- a/scripts/build/internals.sh   ÂMon Dec 06 22:42:07 2010 -0800
>>> +++ b/scripts/build/internals.sh   ÂTue Dec 07 00:56:34 2010 -0800
>>> @@ -20,12 +20,12 @@
>>> Â Â Â Â Âesac
>>> Â Â Â Â ÂCT_DoLog INFO "Stripping all toolchain executables"
>>> Â Â Â Â ÂCT_Pushd "${CT_PREFIX_DIR}"
>>> - Â Â Â Âfor t in ar as c++ c++filt cpp dlltool dllwrap g++ gcc gcc-${CT_CC_VERSION} gcov gprof ld nm objcopy objdump ranlib readelf size strings strip addr2line windmc windres; do
>>> + Â Â Â Âfor t in ar as c++ c++filt cpp dlltool dllwrap g++ gcc gcc-${CT_CC_REAL_VERSION} gcov gprof ld nm objcopy objdump ranlib readelf size strings strip addr2line windmc windres; do
>>> Â Â Â Â Â Â Â[ -x bin/${CT_TARGET}-${t}${CT_HOST_SUFFIX} ] && ${CT_HOST}-strip ${strip_args} bin/${CT_TARGET}-${t}${CT_HOST_SUFFIX}
>>> Â Â Â Â Â Â Â[ -x ${CT_TARGET}/bin/${t}${CT_HOST_SUFFIX} ] && ${CT_HOST}-strip ${strip_args} ${CT_TARGET}/bin/${t}${CT_HOST_SUFFIX}
>>> Â Â Â Â Âdone
>>> Â Â Â Â ÂCT_Popd
>>> - Â Â Â ÂCT_Pushd "${CT_PREFIX_DIR}/libexec/gcc/${CT_TARGET}/${CT_CC_VERSION}"
>>> + Â Â Â ÂCT_Pushd "${CT_PREFIX_DIR}/libexec/gcc/${CT_TARGET}/${CT_CC_REAL_VERSION}"
>>> Â Â Â Â Âfor t in cc1 cc1plus collect2; do
>>> Â Â Â Â Â Â Â[ -x ${t}${CT_HOST_SUFFIX} ] && ${CT_HOST}-strip ${strip_args} ${t}${CT_HOST_SUFFIX}
>>> Â Â Â Â Âdone
>>
>> This will no longer apply, I've reworked that stuff in my tree, as some
>> stuff was missing in there. I'll push that a bit later in the week, sorry
>> for that... :-/
>
> Go, go gadget mq. I'll "hg qpop -a && hg pull -u", fix the other
> problems, and re-post.
>
>>> diff -r 73795525c82e -r 8f7cc140fe23 scripts/crosstool-NG.sh.in
>>> --- a/scripts/crosstool-NG.sh.in   ÂMon Dec 06 22:42:07 2010 -0800
>>> +++ b/scripts/crosstool-NG.sh.in   ÂTue Dec 07 00:56:34 2010 -0800
>>> @@ -151,6 +151,16 @@
>>> ÂCT_CC_CORE_EXTRA_CONFIG="${CT_ARCH_CC_CORE_EXTRA_CONFIG} ${CT_CC_CORE_EXTRA_CONFIG}"
>>> ÂCT_CC_EXTRA_CONFIG="${CT_ARCH_CC_EXTRA_CONFIG} ${CT_CC_EXTRA_CONFIG}"
>>>
>>> +# If we are using a gcc snapshot, get the real "version number" straight
>>> +# here so we don't have to modify the build scripts.
>>> +if [ "${CT_CC_V_4_5_SNAPSHOT}" = "y" -o "${CT_CC_V_4_6_SNAPSHOT}" = "y" ]; then
>>> + Â Âif [ -n "${CT_CC_SNAPSHOT}" ]; then
>>> + Â Â Â ÂCT_CC_VERSION="${CT_CC_VERSION}${CT_CC_SNAPSHOT}"
>>> + Â Âelse
>>> + Â Â Â ÂCT_Abort "You forgot to set the gcc snapshot date!"
>>> + Â Âfi
>>> +fi
>>
>> This should go to the gcc.sh script. I know you needed it to be globally
>> available for the strip phase above, but that will no longer be needed.
>
> ack.

Actually, come to think about it, when you use the snapshot, the
${CT_CC_VERSION} - which for a snapshot would be something like:
4.6-20101204 - would not be what gcc-${CT_CC_VERSION} is.

Say I used said snapshot, gcc-${CT_CC_VERSION} would be 4.6.0, because
that is what is in ${CT_SRC_DIR}/gcc-${CT_CC_VERSION}/gcc/BASE-VER

>>> Â# Compute the working directories names
>>> ÂCT_TARBALLS_DIR="${CT_WORK_DIR}/tarballs"
>>> ÂCT_SRC_DIR="${CT_WORK_DIR}/src"
>>> @@ -572,6 +582,11 @@
>>> Â Â Âfi
>>> Âfi
>>>
>>> +# When referencing the real gcc version, and not the (possibly) snapshot
>>> +# version, use ${CT_CC_REAL_VERSION} instead of the (possibly) snapshot
>>> +# version.
>>> +CT_CC_REAL_VERSION=$(cat ${CT_SRC_DIR}/gcc-${CT_CC_VERSION}/gcc/BASE-VER)
>>> +
>>> Â# Now for the job by itself. Go have a coffee!
>>> Âif [ "${CT_ONLY_DOWNLOAD}" != "y" -a "${CT_ONLY_EXTRACT}" != "y" ]; then
>>> Â Â Â# Because of CT_RESTART, this becomes quite complex
>>
>> Ditto.
>
> ack.

... and because of that, I will still need this CT_CC_REAL_VERSION.

and set gcc-${CT_CC_REAL_VERSION} in internals.sh.

>> Regards,
>> Yann E. MORIN.
>>
>> --
>> .-----------------.--------------------.------------------.--------------------.
>> | ÂYann E. MORIN Â| Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
>> | +33 662 376 056 | Software ÂDesigner | \ / CAMPAIGN Â Â | Â___ Â Â Â Â Â Â Â |
>> | +33 223 225 172 `------------.-------: ÂX ÂAGAINST Â Â Â| Â\e/ ÂThere is no Â|
>> | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL  Â|  v  conspiracy. Â|
>> '------------------------------^-------^------------------^--------------------'
>>
>>
>>
>
>
> -Bryan
>

-Bryan

--
For unsubscribe information see http://sourceware.org/lists.html#faq