diff mbox

gnu: gcc-4.9: Update to 4.9.4.

Message ID 7t4m42uque.fsf@gmail.com
State New
Headers show

Commit Message

Carlos Sánchez de La Lama Oct. 24, 2016, 7:51 a.m. UTC
gcc-4.9.3 has a bug in long double isinf builtin on PowerPC, which
affects glibc versions >= 2.23.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70117

* gnu/packages/gcc.scm (gcc-4.9): Update to 4.9.4.
---
 gnu/packages/gcc.scm | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Leo Famulari Oct. 24, 2016, 1:16 p.m. UTC | #1
On Mon, Oct 24, 2016 at 09:51:53AM +0200, Carlos Sánchez de La Lama wrote:
> gcc-4.9.3 has a bug in long double isinf builtin on PowerPC, which
> affects glibc versions >= 2.23.
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70117
> 
> * gnu/packages/gcc.scm (gcc-4.9): Update to 4.9.4.

Is this different from what we have on core-updates?

http://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/gcc.scm?h=core-updates#n339
Carlos Sánchez de La Lama Oct. 24, 2016, 2:35 p.m. UTC | #2
> On Mon, Oct 24, 2016 at 09:51:53AM +0200, Carlos Sánchez de La Lama wrote:
>> gcc-4.9.3 has a bug in long double isinf builtin on PowerPC, which
>> affects glibc versions >= 2.23.
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70117
>> 
>> * gnu/packages/gcc.scm (gcc-4.9): Update to 4.9.4.
>
> Is this different from what we have on core-updates?
>
> http://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/gcc.scm?h=core-updates#n339

Not at all. I am not tracking core-updates, so I had missed that. I
will check there in the future. Sorry about that.

BTW, is there described somewhere which kind of updates go to
core-updates first? Does everything go first to core-updates?

I am trying to understand to decide on whether I should use cote-updates
as a basis for my contributions instead of master.

Thanks!

Carlos
Marius Bakke Oct. 24, 2016, 5 p.m. UTC | #3
Carlos Sánchez de La Lama <csanchezdll@gmail.com> writes:

>
> BTW, is there described somewhere which kind of updates go to
> core-updates first? Does everything go first to core-updates?

Only packages with a large impact in terms of rebuilds (as calculated by
`guix refresh -l <package-name>`). Obviously gcc is one such package.

Soon there will also be a "staging" branch for packages with a lesser
impact, but still large enough that it's worth building and testing them
on Hydra before merging to master.

See this post for more information:

http://lists.gnu.org/archive/html/guix-devel/2016-10/msg00933.html
Mark H Weaver Oct. 24, 2016, 8:07 p.m. UTC | #4
csanchezdll@gmail.com (Carlos Sánchez de La Lama) writes:

>> On Mon, Oct 24, 2016 at 09:51:53AM +0200, Carlos Sánchez de La Lama wrote:
>>> gcc-4.9.3 has a bug in long double isinf builtin on PowerPC, which
>>> affects glibc versions >= 2.23.
>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70117
>>> 
>>> * gnu/packages/gcc.scm (gcc-4.9): Update to 4.9.4.
>>
>> Is this different from what we have on core-updates?
>>
>> http://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/gcc.scm?h=core-updates#n339
>
> Not at all. I am not tracking core-updates, so I had missed that. I
> will check there in the future. Sorry about that.
>
> BTW, is there described somewhere which kind of updates go to
> core-updates first? Does everything go first to core-updates?

Most changes go to 'master' first and are later merged into
'core-updates'.  However, changes that would force a large number of
rebuilds need to be pushed to another branch, to allow our build farm to
rebuild before its merged to 'master'.

'gcc-4.9' is our default compiler, so it would force a rebuild of
_everything_.  Changes like that are precisely what 'core-updates' is
for.

> I am trying to understand to decide on whether I should use cote-updates
> as a basis for my contributions instead of master.

If you are porting to a new architecture, I would definitely recommend
basing your work on 'core-updates', which will likely be merged into
'master' in the next two weeks.  If you need gcc-4.9.4, that's another
reason to base your work on 'core-updates'.

Thanks for your efforts!

      Mark
Carlos Sánchez de La Lama Oct. 25, 2016, 7:03 a.m. UTC | #5
Hi Marius, Mark,

> Only packages with a large impact in terms of rebuilds (as calculated by
> `guix refresh -l <package-name>`). Obviously gcc is one such package.

> http://lists.gnu.org/archive/html/guix-devel/2016-10/msg00933.html

Thanks! That is exactly the information I was looking for.

> Most changes go to 'master' first and are later merged into
> 'core-updates'.  However, changes that would force a large number of
> rebuilds need to be pushed to another branch, to allow our build farm to
> rebuild before its merged to 'master'.

Understood, makes a lot of sense.

> If you are porting to a new architecture, I would definitely recommend
> basing your work on 'core-updates', which will likely be merged into
> 'master' in the next two weeks.  If you need gcc-4.9.4, that's another
> reason to base your work on 'core-updates'.

I am (I think) almost there, I expect to be able to succesfully build
hello on my ppc machine soon (it's a slow machine, so iterations take
time). I do not feel like changing to core-updates now (though I would
have been better had I worked there since the beggining). Once I have
bootstrapped, I will review the changes and rebase the "core" ones to
core-updates. I should be there before two weeks' time.

Thanks!

Carlos
diff mbox

Patch

diff --git a/gnu/packages/gcc.scm b/gnu/packages/gcc.scm
index c961c84..6f1b744 100644
--- a/gnu/packages/gcc.scm
+++ b/gnu/packages/gcc.scm
@@ -332,14 +332,14 @@  Go.  It also includes runtime support libraries for these languages.")
 
 (define-public gcc-4.9
   (package (inherit gcc-4.8)
-    (version "4.9.3")
+    (version "4.9.4")
     (source (origin
               (method url-fetch)
               (uri (string-append "mirror://gnu/gcc/gcc-"
                                   version "/gcc-" version ".tar.bz2"))
               (sha256
                (base32
-                "0zmnm00d2a1hsd41g34bhvxzvxisa2l584q3p447bd91lfjv4ci3"))
+                "14l06m7nvcvb0igkbip58x59w3nq6315k6jcz3wr9ch1rn9d44bc"))
               (patches (search-patches "gcc-libvtv-runpath.patch"))))))
 
 (define-public gcc-5