[bilibili] #1698 regression: only the first part of a multi-part video is downloaded
====================================================================================
Since #1698, the feature of automatically downloading all parts of a multi-part video, e.g., https://www.bilibili.com/video/av12043403/, has been dropped. Relevant obsolete code: https://github.com/soimort/you-get/blob/v0.4.750/src/you_get/extractors/bilibili.py#L156-L187.
Prior to #1698 (tested on 7d6f68a1592c7d6a29651bdc7b18eaaca7bce3ac with a [faker patch](https://github.com/soimort/you-get/pull/2026) to bypass 403 problem in the U.S.):
```
$ you-get --debug --url https://www.bilibili.com/video/av12043403/
[DEBUG] get_content: http://www.bilibili.com/video/av12043403/
[DEBUG] get_response: http://www.bilibili.com/video/av12043403/index_1.html
[DEBUG] get_response: http://www.bilibili.com/video/av12043403/index_2.html
[DEBUG] get_content: http://interface.bilibili.com/playurl?&cid=19870741&from=miniplay&player=1&sign=c8c43fe8f5a2e61b93c5ac5231b635aa
[DEBUG] url_info: http://tx.acgvideo.com/3/e5/19870741-1.flv?txTime=1499616187&platform=pc&txSecret=d4c278e2eef35ca0bf98e89870746a0d&oi=2360654860&rate=1280000
Site: bilibili.com
Title: Multi-part sample-1、part1
Type: Flash video (video/x-flv)
Size: 0.01 MiB (15676 Bytes)
Real URLs:
http://tx.acgvideo.com/3/e5/19870741-1.flv?txTime=1499616187&platform=pc&txSecret=d4c278e2eef35ca0bf98e89870746a0d&oi=2360654860&rate=1280000
[DEBUG] get_content: http://interface.bilibili.com/playurl?&cid=19870742&from=miniplay&player=1&sign=cf4871820cdbf5b33a974a8d01614ce0
[DEBUG] url_info: http://tx.acgvideo.com/4/b2/19870742-1.flv?txTime=1499616188&platform=pc&txSecret=b10bef5eb2090463cfea8fdb154cd821&oi=2360654860&rate=1280000
Site: bilibili.com
Title: Multi-part sample-2、part2
Type: Flash video (video/x-flv)
Size: 0.01 MiB (15676 Bytes)
Real URLs:
http://tx.acgvideo.com/4/b2/19870742-1.flv?txTime=1499616188&platform=pc&txSecret=b10bef5eb2090463cfea8fdb154cd821&oi=2360654860&rate=1280000
```
Current `develop` (dbe272a8af363544e7fc45a68aa86045e91d66be):
```
$ you-get --debug --url https://www.bilibili.com/video/av12043403/
[DEBUG] url_locations: http://www.bilibili.com/video/av12043403/
[DEBUG] get_content: http://www.bilibili.com/video/av12043403/
[DEBUG] get_content: http://interface.bilibili.com/playurl?cid=19870741&player=1&quality=3&ts=1499605394&sign=efef26a91b153fe1b1eb4eda12a384fd
[DEBUG] get_content: http://comment.bilibili.com/19870741.xml
site: Bilibili
title: Multi-part sample
stream:
- format: flv
container: flv
size: 0.0 MiB (15676 bytes)
# download-with: you-get --format=flv [URL]
Real URLs:
http://tx.acgvideo.com/3/e5/19870741-1.flv?txTime=1499616196&platform=pc&txSecret=12100f02c1909fbb0ee3337d394b9fe0&oi=2360654860&rate=1280000
```
I'll reimplement this feature when I have time, but please feel free to beat me to it (much appreciated actually).
CC @rosynirvana
If they are to recognize `moretension/duti#20` in a link destination as an issue reference rather than a relative URL, they need to specifically have a parser/transformer on link destinations. They can't just blindly expand each token like `moretension/duti#20` to a string `https://github.com/moretension/duti/pull/20`, or `Link ("",[],[]) [Str "moretension/duti#20"] ("https://github.com/moretension/duti/pull/20","")` in the AST (using Pandoc AST syntax) without context awareness.
Too bad this can't be contributed back.
Well, it's not a bug per CommonMark Spec ([Links](http://spec.commonmark.org/0.27/#links)):
```
$ cmark <<< '[Patch merged by upstream](moretension/duti#20)'
<p><a href="moretension/duti#20">Patch merged by upstream</a></p>
$ cmark-gfm <<< '[Patch merged by upstream](moretension/duti#20)'
<p><a href="moretension/duti#20">Patch merged by upstream</a></p>
```
Although it would be nice if they recognized the issue reference.
https://github.com/moretension/duti/pull/20 merged!
> So it would probably make the patch a longer term fix to just use the unversioned directory. (Also, no one is really using Xcode 7 on 10.11 anymore so MacOSX10.11.sdk wouldn't be the correct path anyway.)
Those are good points and I have updated the patch accordingly. From the upstream POV though, it's still good to recognize each Darwin version rather than having a catch-all clause. In particular, the upstream needs the current macOS version as dep_target, and all the catch-all approaches I'm aware of are ugly: <s>parse `system_profiler` output, parse `sw_vers` output</s> (these two don't work when you cross-compile), or parse + do arithmetic on `host_os` (`darwin*`).
However, from our POV we don't really need to set `-isysroot` and `-mmacosx-version-min` (we have `SDKROOT` set already and we don't cross-compile). So a permanent fix for brew would be removing the `sdkpath` and `dep_target` checks altogether in `aclocal.m4`, and remove the aforementioned compiler opts from `Makefile.in`.