Opacity ignored when generating an image of WIDTH between 1941 & 1959 pixels #5181

Closed
ochriste opened this Issue Oct 5, 2015 · 6 comments

Comments

Projects
None yet
3 participants
@ochriste

ochriste commented Oct 5, 2015

when generating a 24bit image with some opacity, if the image has a WIDTH between 1941 & 1959 pixels, the layers with an opacity value between 0 and 100 are generated with full opacity (100)

the following error message is visible when calling mapserv CGI:

*** BUG ***
In create_bits_image_internal: The expression bits == NULL || (rowstride_bytes % sizeof (uint32_t)) == 0 was false
Set a breakpoint on '_pixman_log_error' to debug

Since GitHub is broken I cant join a file to this issue, but you can get it there:

http://dev.camptocamp.com/files/mapserver/test.map

working query:

mapserv -nh "QUERY_STRING=map=test.map&LAYERS=test,test2&FORMAT=image%2Fpng&TRANSPARENT=TRUE&SERVICE=WMS&SRS=EPSG:21781&VERSION=1.1.1&REQUEST=GetMap&STYLES=&BBOX=508675,142220,543000,165440&WIDTH=1716&HEIGHT=1161" > test_1716.png

mapserv -nh "QUERY_STRING=map=test.map&LAYERS=test,test2&FORMAT=image%2Fpng&TRANSPARENT=TRUE&SERVICE=WMS&SRS=EPSG:21781&VERSION=1.1.1&REQUEST=GetMap&STYLES=&BBOX=508675,142220,543000,165440&WIDTH=1960&HEIGHT=1161" > test_1960.png

non-working query:

mapserv -nh "QUERY_STRING=map=test.map&LAYERS=test,test2&FORMAT=image%2Fpng&TRANSPARENT=TRUE&SERVICE=WMS&SRS=EPSG:21781&VERSION=1.1.1&REQUEST=GetMap&STYLES=&BBOX=508675,142220,543000,165440&WIDTH=1950&HEIGHT=1161" > test_1950.png

Mapserver version 7.0.0

@tbonfort

This comment has been minimized.

Show comment
Hide comment
@tbonfort

tbonfort Oct 5, 2015

Member

please disable pixman support on your build and report back.

Member

tbonfort commented Oct 5, 2015

please disable pixman support on your build and report back.

@ochriste

This comment has been minimized.

Show comment
Hide comment
@ochriste

ochriste Oct 5, 2015

tests with disabled pixman support are conclusive: the problem could not be reproduced

what is pixman used for?

ochriste commented Oct 5, 2015

tests with disabled pixman support are conclusive: the problem could not be reproduced

what is pixman used for?

@ochriste

This comment has been minimized.

Show comment
Hide comment
@ochriste

ochriste Oct 5, 2015

"use pixman as an external dependency to enable use of SSE optimized operations inside the AGG renderer."
what kind of performance gain are we talking here ?

ochriste commented Oct 5, 2015

"use pixman as an external dependency to enable use of SSE optimized operations inside the AGG renderer."
what kind of performance gain are we talking here ?

@tbonfort

This comment has been minimized.

Show comment
Hide comment
@tbonfort

tbonfort Oct 8, 2015

Member

I haven't tested. The functionality is disabled by default and not advertized in the docs as it is still experimental.

Member

tbonfort commented Oct 8, 2015

I haven't tested. The functionality is disabled by default and not advertized in the docs as it is still experimental.

@tbonfort tbonfort closed this Oct 8, 2015

@sebastic

This comment has been minimized.

Show comment
Hide comment
@sebastic

sebastic Oct 12, 2015

Contributor

@tbonfort, can the PIXMAN support (and other experimental features) be more explicitly marked as experimental (e.g. like the experimental features in MapCache)? This will help packagers make better decisions which features to enable or not.

It was enabled in the mapserver Debian package because nothing pointed to its experimental nature. I've disabled it now after this issue was raised in Debian Bug #801573.

Contributor

sebastic commented Oct 12, 2015

@tbonfort, can the PIXMAN support (and other experimental features) be more explicitly marked as experimental (e.g. like the experimental features in MapCache)? This will help packagers make better decisions which features to enable or not.

It was enabled in the mapserver Debian package because nothing pointed to its experimental nature. I've disabled it now after this issue was raised in Debian Bug #801573.

tbonfort added a commit that referenced this issue Oct 12, 2015

@tbonfort

This comment has been minimized.

Show comment
Hide comment
@tbonfort

tbonfort Oct 12, 2015

Member

@sebastic done. Sorry about that.

Member

tbonfort commented Oct 12, 2015

@sebastic done. Sorry about that.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment