On Tue, Aug 15, 2017 at 2:07 AM, George Joseph <gjos...@digium.com> wrote:
> > > On Mon, Aug 14, 2017 at 1:04 PM, Dan Jenkins <dan.jenkin...@gmail.com> > wrote: > >> >> >> On Tue, Aug 8, 2017 at 10:44 PM, George Joseph <gjos...@digium.com> >> wrote: >> >>> >>> >>> On Tue, Aug 8, 2017 at 1:15 PM, George Joseph <gjos...@digium.com> >>> wrote: >>> >>>> The option to use the bundled version of pjproject has been available >>>> since January 2016 and is the only "supported" method of using pjproject >>>> now. It's also the only effective way to troubleshoot since things like >>>> DONT_OPTIMIZE and MALLOC_DEBUG are passed down the the bundled build. >>>> >>>> You can still disable it of course with '--without-pjproject-bundled'. >>>> >>>> There have been comments about needing internet access to build but you >>>> can download the tarball and checksum yourself and place them in a known, >>>> static, location, then use the '--with-externals-cache' option to tell >>>> configure where it is. Same for the precompiled codecs and the DPMA. >>>> >>>> You can also use the PJPROJECT_URL environment variable to change the >>>> downlaod URL to a file:/// (or any) URL that points to the directory where >>>> the tarball and checksum are located. >>>> >>> >>> After some discussion I've added the following... >>> >>> To make building without an internet connection easier, a new >>> ./configure option '--with-download-cache' was added that sets the cache >>> for externals (like pjproject, the codecs and the DPMA), AND the sounds >>> files. It can also be specified as an environment variable named >>> "AST_DOWNLOAD_CACHE". The existing '--with-sounds-cache' option / >>> SOUNDS_CACHE_DIR env variable and '--with-externals-cache' option / >>> EXTERNALS_CACHE_DIR env variable remain and if specified, will override >>> '--with-downloads-cache'. >>> >>> >>> >>> >>>> >>>> Thoughts? >>>> >>>> -- >>>> George Joseph >>>> Digium, Inc. | Software Developer >>>> 445 Jan Davis Drive NW - Huntsville, AL 35806 - US >>>> Check us out at: www.digium.com & www.asterisk.org >>>> >>>> >>> >>> >>> -- >>> George Joseph >>> Digium, Inc. | Software Developer >>> 445 Jan Davis Drive NW - Huntsville, AL 35806 - US >>> Check us out at: www.digium.com & www.asterisk.org >>> >>> >>> -- >>> _____________________________________________________________________ >>> -- Bandwidth and Colocation Provided by http://www.api-digital.com -- >>> >>> asterisk-dev mailing list >>> To UNSUBSCRIBE or update options visit: >>> http://lists.digium.com/mailman/listinfo/asterisk-dev >>> >> >> >> Is this not the problem that people didn't like about PJSIP back at the >> very beginning? I know the code isn't inside the Asterisk source code etc. >> But should we be forcing people by default to use bundled? >> > > Yes. But let me tell you why. :) > > It's already bad enough when folks report crashes with no backtraces or > backtraces without DONT_OPTIMIZE and BETTER_BACKTRACES, with 10000 line > dialplans and 2mb log files and say "yeah it only happens every 1000 calls > or so and we can't reproduce it". Now throw in an external pjproject, at > god knows what version, with god knows what compile options, and > symbols???, good luck. It's even worse if they're using a packaged > pjproject because then even the reporter has no idea what compile options > were used or what packager-specific patches were applied and if they didn't > install the symbols, backtraces will be useless. It can take us days or > even _weeks_ to get the reporter to the right configuration for a usable > backtrace. We just can't support that variability any more. With > bundled, we know EXACTLY what they've got and how its configured. *That* > we can support. Also consider that many of the issues with pjproject are > actually memory corruption issues that probably aren't even caused by > pjproject. With the bundled version, when you specify MALLOC_DEBUG in the > asterisk compile options, we turn it on for pjproject as well. That's > something that CAN'T be done with external pjproject, even if you compile > and build it yourself, because it's asterisk-specific code that gets > compiled into the pjproject memory management stuff. > > Given all that, and the fact that bundled is what we test and we patch > bundled and send those patches upstream, I don't think it's unreasonable to > make it the default in 15/master and say "We can't help you troubleshoot > beyond basics if you aren't using the bundled version of pjproject". Now > we can't be assholes about it but we should be firm and if they have an > issue with bundled, *at least tell us why so we can fix it*. I can also > understand packagers (or pbx distros) creating matching asterisk and > pjproject packages and that's fine. Or would be if they actually took > responsibility for providing support, but they rarely do. > > If someone doesn't want to use it, fine. They can specify > --without-pjproject-bundled on the configure command line. They just can't > expect support from us. If I had my way, I'd add a field to the Jira new > issue dialog that makes you confirm that you're using bundled if any > component selected is pjsip related and another one that make you confirm > that, in the event of a crash, you've compiled asterisk with DONT_OPTIMIZE, > BETTER_BACKTRACES and MALLOC_DEBUG and that you've generated backtraces > with ast_coredumper. Don't check the boxes and we don't even look at the > issue. > > Sorry for the rant but I've had a few of these just in the past few weeks. > :) > > >> I personally don't have an issue with it (As I use bundled all the time >> now) - just thinking back to what people had an issue with however many >> years ago >> > > Are you thinking back to issues with bundled or pjproject in general? Or > even back to Asterisk 11 where we had a version of pjproject checked into > the source tree that we used for ICE, etc.? I do remember issues from 13 > but before bundled where we got complaints that we weren't supporting the > "latest" pjproject version but that should be a non-issue now. > > >> -- >> _____________________________________________________________________ >> -- Bandwidth and Colocation Provided by http://www.api-digital.com -- >> >> asterisk-dev mailing list >> To UNSUBSCRIBE or update options visit: >> http://lists.digium.com/mailman/listinfo/asterisk-dev >> > > > > -- > George Joseph > Digium, Inc. | Software Developer > 445 Jan Davis Drive NW - Huntsville, AL 35806 - US > Check us out at: www.digium.com & www.asterisk.org > > > -- > _____________________________________________________________________ > -- Bandwidth and Colocation Provided by http://www.api-digital.com -- > > asterisk-dev mailing list > To UNSUBSCRIBE or update options visit: > http://lists.digium.com/mailman/listinfo/asterisk-dev I was referring to 11 where it was within the source code itself - I know this isn't the same - distro creators can disable it etc etc, can build things with pjsip how they want whereas they couldnt before... I'm more playing devils advocate - I personally don't have an issue with it myself. Just replies had been sparse and I think the above rant is important to be seen in public :)
-- _____________________________________________________________________ -- Bandwidth and Colocation Provided by http://www.api-digital.com -- asterisk-dev mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-dev
