Pierre Labastie wrote:
On 16/11/2016 09:31, [email protected] wrote:
14. Nov 2016 20:35 by [email protected] <mailto:[email protected]>:



     On Sun, Nov 13, 2016 at 5:44 PM, Bruce Dubbs <[email protected]
     <mailto:[email protected]>> wrote:

         Douglas R. Reno wrote:

             On Sun, Nov 13, 2016 at 4:07 PM, Ken Moffat
             <[email protected] <mailto:[email protected]>> wrote:

                 On Sun, Nov 13, 2016 at 12:55:01PM -0600, DJ Lucas wrote:

                                 On Sat, Nov 12, 2016 at 06:56:08PM +0100,
                                 Pierre Labastie wrote:


                         OK, I think I found the culprit: went back to
                         ICU-57.1, compiled FF
                         against it, and no segfaults for about one hour...

                         I'll try to find where the ICU API changed.
                         Pierre


                     All over the place, 58 has not been pretty for many 
packages.
                     This might be a good start:
                     https://bugzilla.mozilla.org/show_bug.cgi?id=1299615


                 For the overall changes,
                 http://source.icu-project.org/repos/icu/icu/tags/release-58-
                 1/APIChangeReport.html

                 But like DJ says, it ain't pretty. The page linking to that,
                 http://site.icu-project.org/download/58#TOC-Common-Changes
                 summariese these as

                 API Changes since ICU4C 57

                 ·   The API change report looks large and noisy this time. 
Most of
                      the changes are from the following:
                      · The layout engine code was removed, resulting in a long
                 list
                        of Removed API.
                      · C++ APIs that used const StringPiece& now pass/return
                        StringPiece by value. All affected functions show as 
being
                        Removed with their old signature and as Added with
                 their new
                        signature. (#12140)
                      · Most COUNT and LIMIT enum constants have been
                 deprecated. (#12420)


             I'm rebuilding Firefox, Seamonkey, and Thunderbird with ICU-57.1 
as we
             speak. Probably about 6 hours until I have any news.

             If this fixes it, do we need to revert to 57.1?


         I would say yes, the revert is needed.   We need to keep an eye on it
         though.  If it bothers enough upstream upstream developers, then thy
         will probably create some compatibility functions.

         Usually in this situation though. old APIs are marked deprecated for a
         while (years) before they are removed.

The API whose change prevents libreoffice to build was announced. There is a
fix in the book.

The segfaults with the mozilla apps is due to a regression. This kind of
things happen... There is a patch which allows firefox and thunderbird to run.
I have to build seamonkey (never done that before), to check it works.

For chromium, internal ICU is used in the book. For sgdisk, ICU should not be
needed as of version 1.0.1 (see ticket #8527).

So, I propose to revert back to 58.1 (with the patch of course). Would it be
OK? Note that apps compiled with 57.1 seem to work with the 58.1 library, at
least FF and TB. No need to recompile AFAICT.

Redoing version 58 is fine. The purpose of -dev is to get things in and working and occasional breakage is expected.

  -- Bruce


--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to