[zanata-users] [Fwd: NetworkManager and Zanata woes]

Fri Feb 12 04:52:42 UTC 2016

Hi Lubo,

(Comments are below.)

On 11 February 2016 at 20:11, Lubomir Rintel <lkundrak at v3.sk> wrote:

> Hello.
>
> It never says on your site that the list is subscriber-only. :(
>
> I can't see a reason why it should be either; I'm probably not
> interested in stuff that's not Cc-ed to me.
>
> I'm wondering if you could forward the message for me?
>

Sorry about the mailing list troubles. We don't allow non-subscribers to
post to the lists because it helps to minimise spam. I've added
zanata-users to the CC, since this isn't really a Zanata development
question. Note that you won't be able to "reply all" unless you join the
zanata-users list.

>
> Thank you,
> Lubo
>
> -------- Forwarded Message --------
> From: zanata-devel-owner at redhat.com
> To: lkundrak at v3.sk
> Subject: NetworkManager and Zanata woes
> Date: Thu, 11 Feb 2016 05:08:57 -0500
> Message-id: <mailman.10055.1455185337.27305.zanata-devel at redhat.com>
>
> You are not allowed to post to this mailing list, and your message has
> been automatically rejected.  If you think that your messages are
> being rejected in error, contact the mailing list owner at
> zanata-devel-owner at redhat.com.
>
>
> ---------- Forwarded message ----------
> From: Lubomir Rintel <lkundrak at v3.sk>
> To: zanata-devel at redhat.com
> Cc:
> Date: Thu, 11 Feb 2016 11:08:45 +0100
> Subject: NetworkManager and Zanata woes
> Hi,
>
> I've tried to use Zanata for NetworkManager. So far my experience has
> been rather frustrating. I'm writing in hope that you'd be able and
> willing to help me, since I'm still convinced Zanata is the right tool
> for the job.
>
> So, here's what bothers me:
>
> 0.) I had trouble finding out where to report problems. This is in
> fact, why I'm writing a mailing list post instead.
>
> There's apparently a bugzilla.redhat.com product a JIRA instance and
> GitHub issues. I'd prefer to use the Bugzilla, since I find it most
> convenient, but I suspect you've deprecated it as it's not in the list
> when I click "New".
>

It should be our JIRA: https://zanata.atlassian.net/ We migrated all our
bugzilla issues to JIRA a while back.

I'm not sure which Zanata repo has GitHub Issues enabled, but we should
probably turn it off. Please let us know!

>
> 1.) What's the difference between fedora.zanata.org and
> translate.zanata.org?
>
> I've added my project to the former, follwing the Fedora documentation,
> only to find out OpenID doesn't work. I should probably have used the
> latter, but why does the former exist then? I couldn't find out.
>

fedora.zanata.org is basically for Fedora projects - only Fedora
contributors can log in and submit content or translations.  (Think
git.fedorahosted.org.)

translate.jboss.org is intended for jboss.org projects.

translate.zanata.org is intended for any open source project.
(Think GitHub.)

When you say OpenID doesn't work, are you saying you couldn't log in to
fedora.zanata.org with a Fedora account?

>
> 2.) Presumably translations are often lost or duplicated.
>
> I've upload my project to the fedora.zanata.org and
> translate.zanata.org. Both show different percentage on all languages.
>
> This really freaks me out. I don't want to loose our translators work.
>

Well, if the same project is uploaded to multiple instances, it is possible
translators will find different instances and submit translations all over
the place. I strongly recommend you make all but one of them read-only, and
link to your choice of canonical instance.

Another reason the stats could be different is the upload problems you've
been having (partly because of the Python client).

>
> 3.) The downloads don't work
>
> I thought I'll download one translation (I chose "Asamese" language)
> and compare them to see what's different, but after an hour it's stuck
> on "Download as translation files with "This may take a few minutes to
> complete. Processing 1 of 2 50%"".
>

How were you trying to download? Python client? Web UI?

>
> 4.) Everything is in general too slow; sometimes too slow to be
> actually usable.
>
> Perhaps it's that the project is too big?
>
> The initial upload takes many hours to finish and I needed to run it in
> busy loop since it's almost guaranteed to fail with a timeout at some
> point.
>

Uploading can be slow, but it should be more reliable with the Java client.

>
> Also, loading [1] takes 36 seconds for me.
>
> [1]
> https://translate.zanata.org/zanata/iteration/view/NetworkManager/master
>
>
Yes, unfortunately that page has performance problems with the stats
gathering. We're definitely hoping to improve that in future versions.

> 5.) Python client seems somewhat buggy
>
> [lkundrak at belphegor NetworkManager]$ zanata version create
> [INFO] Loading zanata project config from:
> /home/lkundrak/src/NetworkManager/zanata.xml
> [INFO] Loading zanata user config from: /home/lkundrak/.config/zanata.ini
> [INFO] zanata server: https://fedora.zanata.org
> [INFO] zanata python client version: UNKNOWN, zanata server API version:
> 3.8.3
>  Error 404 - The requested resource/project is not available
> Please take a reference in
> https://github.com/zanata/zanata/wiki/Python-Client-Troubleshooting
> [lkundrak at belphegor NetworkManager]$
>
> In fact the version create failed because the version is not there yet.
> But there's no way I could tell this is a client bug; the error
> handling seems totally broken. The only thing I got was a
> broken/unrelated link.
>

(Which I'm guessing should probably point to
https://github.com/zanata/zanata-python-client/wiki/Troubleshooting )

The Python client isn't really supported by the core Zanata team, and I
don't recommend using it, partly because it tends to miss out on features
which are added to the Java client first, eg features which help with
larger file uploads.

>
> 6.) There's no option to enable tracing in client
>
> And the error messages are somewhat cryptic to the point the raw HTTP
> response would be a lot helpful:
>
> "error: ''"
>
> "Exception while decoding No JSON object could be decoded its may due
> to file already exists on the server or not a PO file"
>
> Turns out, the latter is because this is in the actual response body:
>
> warning: Could not find TextFlow for TextFlowTarget
> dd139eab964f6527960f27dad31dfe48 with contents: []
> warning: Could not find TextFlow for TextFlowTarget
> 875069b2bcfe1886deff764cb09e4f2c with contents: [ବ୍ୟବହାର ବିଧି: nmcli
> connection add { ARGUMENTS | help }
> ...
>
> Not sure if it should bother me, but it does given I suspect some
> translations get lost.
>

That just means those translations correspond to English strings which
aren't in the POT file.  This is usually because they are translations of
obsolete strings. In fact, if you were to run msgmerge, they should be
marked as obsolete. On the other hand, if you get hundreds and hundreds of
them, it may be a sign that you have the wrong PO file for that POT file.

>
> So here I stand, confused, frustrated and worried about broken
> translations. I'd hate to give up attempts to use Zanata since the
> translator experience is good, but I completely rely on your help now.
>
> Perhaps some of you, the Zanata developers, could give importing
> NetworkManager translations a try and see what goes wrong. It shouldn't
> be too different from other gettext projects; just clone the git repo,
> configure it and do a "make -C po NetworkManager.pot" to create the
> template file.
>
>
I tracked down the git repo here: git://
anongit.freedesktop.org/NetworkManager/NetworkManager

I ran autogen.sh, but I couldn't find all the rpms it wanted - I gave up
when I got to 'nss'. So I converted the latest PO I could find into a POT
file (gd.po), and I was able to upload the POT and PO file to a test server
successfully, so I don't think Zanata is having any trouble parsing your
file.

I think your upload problem (with the two minute proxy timeout) is because
the Python client doesn't support batch uploads for translation files,
perhaps combined with general slowness on the server side of things.

BTW, the Java client has the HTTP tracing option you wanted:  --log-http.

I would encourage you to report the problems you have been seeing in the
issue tracker: https://zanata.atlassian.net/

Thanks for the feedback, Lubo.

> Thank you,
> Lubo
>

Regards,

Sean.

-- 
Sean Flanigan

Principal Software Engineer
Globalisation Tools Engineering
Red Hat
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/zanata-users/attachments/20160212/56cb0c90/attachment.htm>