mantohtml is run during build. Thus it needs to be built with the build
architecture compiler (or execution fails). The obvious part is switching to
CC_FOR_BUILD. That also depends on it not requiring any other cups components.
In particular, removing uses of strlcpy and replacing host architecture-
specific includes is thus needed.
Bug-Debian: https://bugs.debian.org/837936
Forwarded: not-needed
When running it without arguments it is supposed to read the local CUPS's
cupsd.conf and show a summary of the setting. in CUPS 2.3.1 it shows a mess
with a lot of HTML inside and this is due to the fact that when loading the
file via HTTP using the /admin/cups/cupsd.conf path the scheduler calls the
admin.cgi program which returns the admin front page of the web admin
interface. cupsctl then tries to interpret that as the config file and displays
garbage. Even worse is if you run cupsctl with command line argument (one of
the five switches or a key=value pair) to change a setting. It seems to load
cupsd.conf again and gets again the HTML code of the web interface page.
cupsctl tries to interpret this again, producing garbage, adds the
user-supplied setting and writes all this back into cupsd.conf. Then it tries
to restart the scheduler which fails due to the broken config file.
The problem is that in the file scheduler/client.conf, in the function
get_file() the URI from the client is at first checked whether it begins with
"/admin/" and in this case the CGI program admin.cgi is responsible. Only after
that the check for "/admin/conf/cupsd.conf" comes and is never reached.
I have changed the order now appropriately and this way cupsctl works again.
Note that the problem only occurs if the web interface is active and the
cupsctl command is issued by a non-root user.
This is a regression caused by issue #5652.
Bug: https://github.com/apple/cups/issues/5744
Gbp-Pq: Name 0005-Fix-scheduler-cupsd.conf-load.patch
mantohtml is run during build. Thus it needs to be built with the build
architecture compiler (or execution fails). The obvious part is switching to
CC_FOR_BUILD. That also depends on it not requiring any other cups components.
In particular, removing uses of strlcpy and replacing host architecture-
specific includes is thus needed.
Bug-Debian: https://bugs.debian.org/837936
Patch-Name Build-mantohtml-with-the-build-architecture-compiler.patch
Gbp-Pq: Name 0033-Build-mantohtml-with-the-build-architecture-compiler.patch
It it is less confusing and a better default on most machines.
Amend documentation accordingly.
Author: Didier Raboud <odyx@debian.org>
Origin: vendor
Patch-Name: cups-set-default-error-policy-retry-job.patch
Gbp-Pq: Name 0030-Set-default-job-error-policy-to-retry-job.patch
managing print queues broadcasted from other servers by default. Many
networks do not have valid host names for all machines.
Bug-Ubuntu: https://bugs.launchpad.net/bugs/449586
Patch-Name: do-not-broadcast-with-hostnames.patch
Gbp-Pq: Name 0023-Do-not-use-host-names-for-broadcasting-print-queues-.patch
… of HP and Ricoh to obtain the device IDs of network-connected
printers. This way we get more reliable information about make and model
and in addition the supported page description languages, which allows one to
identify whether an optional PostScript add-on is installed or for an
unsupported printer which generic PPD is the best choice (requested by
Ricoh).
Bug: https://github.com/apple/cups/issues/3552
Patch-Name: Let-snmp-backend-also-use-manufacturer-specific-MIBs.patch
Gbp-Pq: Name 0019-Let-snmp-backend-also-use-manufacturer-specific-MIBs.patch
Revert commit 830cfed95a5349b008999eebd34f5233bc35829c
"Update "make test" to include alternate filter."
.
This fixes the error lines counting in the test, and is unneeded as pdftourf
is not shipped in Debian.
Origin: upstream
Patch-Name: tests-no-pdftourf.patch
Gbp-Pq: Name 0016-Tests-Do-not-test-pdftourf.patch
This should make the test-suite work in pbuilder. It apparently also fixes an
ipp backend problem in the test-suite.
Origin: vendor
Patch-Name: tests-use-ipv4-lo-address.patch
Gbp-Pq: Name 0014-Tests-Use-127.0.0.1-instead-of-localhost-to-help-pbu.patch
Permits building in non-"C" environments.
Origin: vendor
Patch-Name: tests-fix-ppdLocalize-on-unclean-env.patch
Gbp-Pq: Name 0013-Tests-Force-LC_-environment-variables-when-testing-n.patch
running a new test. Fixes failures on slow architectures like armel and
mipsel.
Bug-Debian: https://bugs.debian.org/670878
Patch-Name: tests-wait-on-unfinished-jobs-everytime.patch
Gbp-Pq: Name 0012-Tests-Make-sure-that-all-scheduled-jobs-are-finished.patch
Ignore the 'Unable to write uncompressed print data: Broken pipe' error that
sometimes breaks the errorlines counting on kfreebsd for some reason.
Origin: vendor
Patch-Name: tests-ignore-kfreebsd-unable-to-write-uncompressed-print-data.patch
Gbp-Pq: Name 0009-Tests-ignore-the-failure-to-write-uncompressed-data.patch
Ignore the 'Address family not supported by protocol' error that breaks the
errorlines counting on buildds without a working IPv6 stack.
Origin: vendor
Patch-Name: tests-ignore-ipv6-address-family-not-supported.patch
Gbp-Pq: Name 0008-Tests-ignore-errors-triggered-on-ipv6-deprived-hosts.patch
Ignore the 'loadFile failed: temp file: not a PDF file' error that sometimes
breaks the errorlines counting on kfreebsd-amd64 for some reason
Origin: vendor
Patch-Name: tests-ignore-kfreebsd-amd64-not-a-pdf.patch
Gbp-Pq: Name 0007-Tests-ignore-loadFile-failures.patch
Ignore the following errors:
"[cups-deviced] PID * (usb) crashed on signal 11!" and
"[cups-deviced] PID * (dnssd) stopped with status 1"
They seem to regularly happen on Debian/Ubuntu buildds and break error lines
counting.
Patch-Name: tests-ignore-usb-crash.patch
Gbp-Pq: Name 0006-Tests-ignore-usb-dnssd-backend-unexpected-exits.patch
These warnings change the expected number of warnings
Patch-Name: tests-ignore-warnings.patch
Gbp-Pq: Name 0005-Tests-Ignore-warnings-from-colord-and-Avahi.patch
When I create a CUPS queue with a PPD file with choice names "Tray-1", "Tray-2",
... in the InputSlot option CUPS translates these names to double-dashed IPP
attribute names: "tray--1", "tray--2", ... in the "media-source" attribute, both
when passing a job to the printer with the IPP backend, making the printer
ignore the tray choice, and also when answering a get-printer-attributes IPP
request from a client. This happens when in the PPD a dash is followed by a
digit, as the pwg_unppdize_name() function in cups/ppd-cache.c inserts a dash
whenever a non-digit is followed by a digit in the PPD name. As IPP attribute
names generally never have double-dashes and also no dashes in the beginning or
the end of the name, I have modified the pwg_unppdize_name() function
appropriately.
Bug: https://github.com/apple/cups/issues/5740
Bug-Debian: https://bugs.debian.org/949315
Gbp-Pq: Name 0004-Fix-conversion-of-PPD-InputSlot-choice-names.patch
This is needed:
- to comply with Debian Policy
- because it is both nonsensical to not do so
- it also breaks system checkers, bug reporting, etc
Bug: https://github.com/apple/cups/issues/2935
Bug-Debian: https://bugs.debian.org/410171
Patch-Name install-root-backends-world-readable.patch
Gbp-Pq: Name 0002-Install-root-backends-world-readable.patch
Automatically generated using sed -i -s 's/^-/\\-/g;s/ -/ \\-/g;s/"-/"\\-/g' man/*.man*
Also fix a missing backslash escape
Origin: vendor
Bug-Upstream: https://github.com/apple/cups/issues/4299
Patch-Name: manpage-hyphen-minus.patch
Gbp-Pq: Name 0001-Fix-hyphen-used-as-minus-sign-manpage-errors.patch
Didier Raboud