Le 5 déc. 11 à 17:09, Bjoern Michaelsen a écrit :
Hi Eric, all,
Hi Bjoern,
I can not leave this standing as is.
Same for me : it was my time to write the code after all, and in
return, I feel bad just because I do not see the minimalistic "thank
you".
Well, the people who know me know I'm always glad and happy to
share my code, everytime I can. But this is not the first time :
previous one was about the ARM assembler optimisation in
interlock.c for armv7 + ISA . It was "fixed" (means my name
added) after a (friendly) discussion with Jani Monoses and Björn
Michaelsen.
You are implying there was a "first time" were LibreOffice cleared
your name of your contribtion.
Exactly. Whatever the origin (Debian or Mars commiter or ...) it was
commited "as it" in LibreOffice.
This is simply not true.
Please read again : I wrote the issue was solved / fixed by you and
Jani Monoses. Said differently "Thanks to both of you ". If you read
something else, I'm sorry, but you misunderstood me.
The ARM patch was upstreamed from Debian Packaging, where it was
already cleared of any author information or bug reference.
It had therefore to be assumed to be of Debian origin. If you want
to put any blame, but it there, not on LibreOffice.
Was the code commited or not in LibreOffice at the end ? The answer
is : yes, it was.
So either the guy who commited the changes is responsible, or there
is nobody who did the commit, and code without clear origin is
included (what is basicaly not good)
Once the truth came to light (IIRC by my Canonical co-worker Jani
Monoses hinting you at it, which does not exactly suggest malicious
intend),
That's what I wrote : "friendly".
we very quickly contacted you to rectify the situation.
... and it was solved, I do not contest.
Ironically, in the process of this, we discovered that recent gcc
versions provide better code themselves, so none of this code is
even used in current releases.
That's not the point.
As the original ASM code is not even SMP-safe
When I wrote it, was Cortex A8 time, means mono-core processor, so, I
didn't think imediately to add data memory barrier (aka dmb) , indeed.
FYI OOo4Kids svn log will show I wrote "thanks to Jani Monoses for
his contribution for SMP case" (or something very close).
and all the compilers Ubuntu use provided a better solution, I
could have dumped that patch -- but we left it in for vanilla gcc
4.5 versions (not Ubuntus fixed one). And we (actually Jani) fixed
the original patch, so that it would be SMP-safe on old, non-Ubuntu
compilers. Looking back, I would have loved to have never touched
that patch at all: it created a lot of work and no extra
value for LibreOffice on Ubuntu.
Every code has a life time. when it was written, it was usefull (and
to my eyes it will be a while on some ARM machines). But again, this
is not the point.
As for the patches lifted from Apache by Thorsten: Complain to the
guys who maintain that nostalgic SCM at Apache to provide more
explicit author information.
The commiter should have read the code, verified the ownership, and
in case of doubt, rectified : as I understood the "spirit" of the
thing, in LibreOffice code, the real name of the author MUST appear
for every commit.
LibreOffice really cant be blamed if Apache publishes your work in
ways that do not suit you.
I have really no problem with Apache, and my motivation is to share
and work on exciting features. I don't care if any of my
contributions end anywhere.
But some basic rules must be respected. Thanks [EOT for me]
Regards,
Eric
--
qɔᴉɹə
Projet OOo4Kids : http://wiki.ooo4kids.org/index.php/Main_Page
L'association EducOOo : http://www.educoo.org
Blog : http://eric.bachard.org/news
