<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<p>Hi Giovanni,<br>
</p>
<br>
<div class="moz-cite-prefix">Le 01/03/2018 à 12:52, Giovanni Manghi
a écrit :<br>
</div>
<blockquote type="cite"
cite="mid:CACfnYnSFTon_HowUaLreZR=nOQXFtn0LPW+FRrKqV_FrfRdPsA@mail.gmail.com">
<pre wrap="">Hi all,
now that 2.18 is LTR I guess that many have also updated their QGIS
Server installation to this release. Problem is that I see a few
issues in 2.18 and would like to have gsome feedback from others.
*) slower: it seems this version is slower compared to 2.14, at least
for some specific request like WFS GetFeatures. See for example:
<a class="moz-txt-link-freetext" href="https://issues.qgis.org/issues/18249#note-1">https://issues.qgis.org/issues/18249#note-1</a>
where in the same condition (same server, same postgis datasource of
around ~25000 polygons) 2.18 is about 30 seconds slower than 2.14.
</pre>
</blockquote>
<br>
This issue is simply due to the transformation of the geometry from
the layer CRS to the GeoJSON CRS standard: EPSG:4326<br>
<br>
<blockquote type="cite"
cite="mid:CACfnYnSFTon_HowUaLreZR=nOQXFtn0LPW+FRrKqV_FrfRdPsA@mail.gmail.com">
<pre wrap="">
*) At some point in 2.18 development -in the print composer,
properties of a legend- the "none" value for the "map" option was
removed. This option (together with "auto update" unchecked) allowed
to do GetPrint requests where a legend always shown the same fixed
entries regardless of the layers being requested. Moreover now the
layers legend (from a GetPrint) frequently shows "?" instead of the
correct legend. Not sure if the two things are related.
</pre>
</blockquote>
<br>
It will be great to have the possibility to allow nullptr in the map
combobox.<br>
Nyall, do you have an idea how to enhance <span
class="blob-code-inner">QgsComposerItemComboBox introdiuced by
<a class="moz-txt-link-freetext" href="https://github.com/qgis/QGIS/commit/1b4bd47076103e931e642c9c2b6a363f14b20a45">https://github.com/qgis/QGIS/commit/1b4bd47076103e931e642c9c2b6a363f14b20a45</a></span><span
class="blob-code-inner"></span> ?<br>
Issue already exists: <a class="moz-txt-link-freetext" href="https://issues.qgis.org/issues/16899">https://issues.qgis.org/issues/16899</a><br>
<br>
<blockquote type="cite"
cite="mid:CACfnYnSFTon_HowUaLreZR=nOQXFtn0LPW+FRrKqV_FrfRdPsA@mail.gmail.com">
<pre wrap="">
*) Crashes on GetPrint: adding some type of elements in print
composers (like html text boxes) caused QGIS Server to crash when
doing a GetPrint *if* the *headless* server didn't had a fake xserver
installed. Now in 2.18 this happens even if QGIS Server is installed
on a Desktop OS, unless installing the fake x server. Not sure is
related but this now affects also Windows machines where a GetPrint
requests (of a layout with one of this problematic elements) make
always QGIS Server crash. This was not the case for 2.14.</pre>
</blockquote>
<br>
The solution is probably to update the environment variable
QT_GRAPHICSSYSTEM to 'raster' has explain
<a class="moz-txt-link-freetext" href="https://issues.qgis.org/issues/15440">https://issues.qgis.org/issues/15440</a> and in a lot of other project:<br>
<a class="moz-txt-link-freetext" href="https://stackoverflow.com/questions/6168570/qgraphicsview-slow-scale-performance-under-linux">https://stackoverflow.com/questions/6168570/qgraphicsview-slow-scale-performance-under-linux</a><br>
<a class="moz-txt-link-freetext" href="https://forum.kde.org/viewtopic.php?t=90821">https://forum.kde.org/viewtopic.php?t=90821</a><br>
<br>
I'll test this solution.<br>
<br>
Regards,<br>
René-Luc<br>
</body>
</html>