Cairo-1.8.8

[Randy McMurchy]

Version increment to 1.8.6

​http://cairographics.org/

[willimm]

Hey, I beat you to opening this ticket! Close as a duplacate of #2702.

[Randy McMurchy]

Ken, I can do this update right now if you like. Let me know.

[Randy McMurchy]

I'm sure this was discussed, however, I don't recall it. Anyway, why is the experimental XCB backend recommended for this package?

Additionally, I'll be adding the Pixman package to the book here in a few minutes as it is now required by Cairo.

Lastly, and I know this was discussed, but I cannot remember it or find the discussion, but why exactly is X required for this package, it will build without it as best as I can tell.

[ken@…]

As to doing the upgrade now, what is the urgency ? Have you found a problem which is in 1.8.4 but already solved in 1.8.6 ?

For xcb, do you want the users to continue living in the "dark ages", or to use what everyone else is using ? I've been using xcb for a very long time, it all works with the LFS-6.4 toolchain.

For the dependency on Xorg itself, I don't recall and I can't see anything relevant in my notes.

[Randy McMurchy]

Hmmm, why do I feel there is a confrontation starting? :-)

Anyway, there is no urgency. However, I have it installed, have tested it, and want to get the book updated. We don't sit on package versions when there is Editor interest to update it. Another thing is that I wanted to fix the dependencies, which are incorrect right now.

As far as your comment about xcb (which was unnecessary by the way), you didn't answer my question. Saying you've been using an experimental portion of some code does not make it "recommended". I was hoping for a bit more dignified answer such as "I'm recommending it because package XYZ is much more robust using it", or something along those lines.

Right now, your answer does not qualify it being a recommended dependency, seeing how the developers still consider it experimental code.

And as far as the X dependecy goes, it is definately not required.

[Randy McMurchy]

I forgot to comment that the package is slated for BLFS-6.4. I cannot see that changing. We are tentatively trying to achieve a March date for the first release candidate. My question is this:

How long to you anticipate waiting until you do this update?

[Randy McMurchy]

Modified milestone from 6.4 to 6.5

[Guy Dalziel]

I don't recommend the use of xcb with Cairo. xcb in Cairo is highly experimental and is only there in order to be previewed. There is a crash in the Cairo testsuite using xcb 0.3.5 (compatibility of previous versions unknown) on the test large-font-xcb-argb32. Are there any objections to the removal of --enable-xcb?

[Randy McMurchy]

As you can see from the messages in this ticket, there was at one time obviously some folks who thought it necessary. You may try to dig through the archives. Probably DJ and Dan Nicholson would have the best input. Too bad Dan isn't around any longer.

I remember something about it being needed for something that I can't remember. I 'm no help at all. I just read the messages in this ticket again and it appears Ken had input as well. As you can see, I was against its recommendation from the beginning.

But that puts us in the "dark ages" :-)

[ken@…]

Apart from the testsuite failure, do you have any justification for "*highly* experimental' ? My memory says that most distributions enable it (but I could be wrong), and the maintainer might just be unnecessarily conservative in his wording.

The wording in the book for test failures was correct last time I tried the tests (many failures, don't recall any crash).

[ken@…]

FWIW, the testsuite for 1.8.8 with xcb-utils-0.3.0 on LFS-6.4 appears to hang after

solid-pattern-cache-stress-svg11-rgb24 [0] (similar) :	PASS

I killed it after 10 minutes, and there were a vast number of failures (55) before that - it hadn't reached the test you quoted.

[Guy Dalziel]

My justification for calling it highly experimental, apart from the fact it crashes, is due in part to that fact that even Mesa doesn't use it by default. We also don't enable it by default for Mesa either. I've seen many bug reports from all over the place showing crashes with Mesa enabled with XCB, and that's with a package in which XCB isn't "experimental". I don't have a problem leaving it in the command explanations for people that want to try it, but I do have a problem with recommending the usage of it.

[ken@…]

OK, I've not seen the crashes, nor any reports.

FWIW, if I don't enable xcb, the testsuite completes. It reports 11 failures, but there were 62 lines starting with 'FAIL' in the log, so I can't guess how many separate _tests_ would be regarded as failing with xcb.

[Guy Dalziel]

I'll move it to optional for now, if you get better results without it than with it then it's the best course of action. I'm not entirely sure what uses XCB in Xorg, but all the latest stuff feels pretty stable. I imagine that the issues of XCB in Cairo are more a case of lack of support than anything. At minimum I'd expect it to pass the testsuite reasonably.

[Guy Dalziel]

I compiled Cairo last night on my pure CLI laptop, I discovered that X is not a requirement. However, libpng and pixman are, along with a font backend such as fontconfig. I shall move X to recommended and add libpng, pixman, and fontconfig to required. While libpng can be disabled it merely forces Cairo to look for another backend, and a basic requirement must be met. Does anyone think that libpng should be optional? While PNG can be replaced with SVG I think such nuances might be better suited for the wiki.

[(none)]

Milestone 6.5 deleted