lp:~naesten/gs-gpl/trunk-gs

Created by Samuel Bronson on 2010-09-21 and last modified on 2011-04-19
Get this branch:
bzr branch lp:~naesten/gs-gpl/trunk-gs

Branch merges

Related bugs

Related blueprints

Branch information

Owner:
Samuel Bronson
Project:
GS-GPL
Status:
Development

Import details

Import Status: Failed

This branch is an import of the Subversion branch from http://svn.ghostscript.com/ghostscript/trunk/gs.

The import has been suspended because it failed 5 or more times in succession.

Last successful import was on 2016-04-20.

Import started on 2016-04-24 on pear and finished on 2016-04-24 taking 30 seconds — see the log
Import started on 2016-04-22 on russkaya and finished on 2016-04-22 taking 30 seconds — see the log
Import started on 2016-04-21 on russkaya and finished on 2016-04-21 taking 30 seconds — see the log
Import started on 2016-04-20 on pear and finished on 2016-04-20 taking 25 seconds — see the log

Whiteboard

20100922 maxb Approved. Perhaps for clarity's sake we should delete the all-in-one lp:~naesten/gs-gpl/trunk ? It doesn't seem like a useful branch to work with.

Recent revisions

9171. By robin on 2011-04-19

Changes to localcluster code to attempt to implement bmpcmphead. This certainly
doesn't break anything, but I don't actually ever seem to get any differences
out when running bmpcmphead.

CLUSTER_UNCHECKED

9170. By robin on 2011-04-19

Fix for bug 692152.

We 'bend' the size of pattern tiles, against the spec, ostensibly to better
match Acrobat. The fix here is simply to avoid such bending in the case where
TilingType is 2. According to the spec, TilingType 2 says specifically not
to use the same device pixel size for every repeat of the pattern, and that
is exactly what we were doing.

This cures the original bug, and produces 464 changes in the cluster tests.
Some (like ps3cet/18-02B.PS) are clear progressions. Others are less obviously
improvements, but comparison with Acrobat (and discussion with Ken/Chris)
shows that it's no worse than before, just 'different'.

9169. By alexcher on 2011-04-18

Fix reading of an embedded ICC profile from DSC comments and a
potential SEGV when the value of /N exceeds 8. Bug 692156.

9168. By mvrhel on 2011-04-18

Further movement toward getting fast thresholding operations working for CMYK output devices. Addition of special op to detect if a device is a planar type, creation of threshold screen for all the planes and spatial resampling of color input to output resolution for portrait case. A commit to get things in trunk before we convert from SVN to git.

9167. By alexcher on 2011-04-18

Consider any xref entry with 0 offset as a free entry. Earlier revisions
rebuilt the xref table when offset == 0 but generation != 0.
However, rebuilding is an unreliable process and should be avoided if
possible. Bug 692159.

9166. By chrisl on 2011-04-18

Add a fallback check for libpaper support.

Older versions of autogen appear to have a bug which causes the AC_CHECK_LIB
macro to mis-identify the libpaper development library as being present when
it is not.

So add a second check check based on the presence of paper.h using
AC_CHECK_HEADER().

No cluster differences expected.

9165. By alexcher on 2011-04-16

Instead of searching for %%EOF marker (which is often damaged) and browsing the
file backwards, search for 'startxref' and take the next token as an xref
position. Besides fixing the reported bug, this greatly simplifies the search
for xref position. Bug 692153.

9164. By mvrhel on 2011-04-14

code to support rendering color source images to a mono device using thresholding. The code is currently disabled pending additional testing.

9163. By mvrhel on 2011-04-13

Reorganization of thresholding code to make it easier to get the threshold operation working for color images as input as well as the case when we go to a cmyk planar device.

9162. By robin on 2011-04-13

Fix reference counting of compositors in clist rendering; previously
the code assumed that it was the only claimant of a compositor and
closed/freed the device regardless of the reference count.

Cluster testing shows no differences.

Branch metadata

Branch format:
Branch format 7
Repository format:
Bazaar repository format 2a (needs bzr 1.16 or later)
This branch contains Public information 
Everyone can see this information.

Subscribers