Dev:Doc/AskUsAnything

提供: wiki
< Dev:Doc
2018年6月29日 (金) 06:13時点におけるYamyam (トーク | 投稿記録)による版 (1版 をインポートしました)
(差分) ← 古い版 | 最新版 (差分) | 新しい版 → (差分)
移動先: 案内、 検索

Developers: Ask Us Anything!

Running Blender project there are so many things we're are asked on a regular basis

(how and why things are...) and why we don't do it different!

While talking to users directly is fine, we often end up repeating ourselves.

So this page is an informal Q&A on common user questions relating to the Blender project.

Of course developers don't always completely agree,
this page can at least answer some common questions on generally accepted practices in the Blender developer community.

See: BlenderArtists Thread to ask questions

NOTE
Questions about specific features in development can't be properly handled on this page.
Planning details change too often and don't fit well with general Q&A.


For more specific development FAQ see: Developer FAQ

Release Cycle

For details on our release cycle, see: Release Cycle

Why not release Blender when it's ready instead of rushing them every few months?

In the years prior to 2.5x development we used this approach, and found it had a number of problems...

From the development side:

  • With active developers maintaining many areas, some volunteers with other commitments - we couldn't always rely on specific tasks being finished in a fixed timeframe.
  • Since Blender's a very complex application, it became increasingly difficult to pick the perfect moment to release for _all_ areas.
    Consider the wide range of areas: rendering, shading languages, modifiers, API's, physics, game engine, file format support, painting, hardware such as tablet/3d-mouse support... the list goes on!
  • The long gaps between releases allowed code to be in an unfinished/broken state for periods of time.
  • With over 6 months between some releases, developers would try to push changes weeks before the release, even when they weren't quite ready...
    (otherwise it may not get into a release for another year).

From the user side:

  • When a bug was fixed, users didn't know how long it would be until that got into a stable release.
    (assuming it wasn't important enough for a bugfix release).
  • Users wouldn't have access to features which were ready for release... because other areas were still in development.

Instead of frantically rushing out releases, we've largely streamlined the process, so it's not a determent to regular development.

To support periodic releases, we strive to keep our main branch (master) always stable, developing features in branches, which are merged into master when ready.

This is not to say our release process is perfect, but for now we're satisfied with the current release cycle and aren't likely to change it significantly.

Release-cycle proposal & mailing list discussion:

Mailing list discussions on this topic (2011, before migration to GIT).

--Ideasman42 21:55, 5 May 2015 (UTC)

Would you consider a long term support release (LTS)?

A: There would need to be enough interest from users and support from developers. While the option to provide different kinds of releases is always open to us, back-porting many changes from newer releases splits up developer resources.

If theres enough demand from users on an older release who can't upgrade, we could always provide releases that only contain critical fixes.

Is the release cycle completely frozen to focus exclusively on solving bugs and enhance existing features?

No, we don't have a period in the release cycle specifically for fixes and enhancements.

Before release (2-3 weeks) we only accept bug fixes, however - we are rather strict about the kinds of bug-fixes we accept, so it's not a good time to work on more general improvements.

How do I find whats been added to daily builds?

A: Work in progress release notes are maintained at the same location as regular release notes, (current version +1).

See top level: Release Notes page.

User Community

How can users help with Blender development

Q: What are some things users can do to help with Blender development?

A: There are various ways users can help out.

  • Report bugs:
    If you run into problems, even minor glitches, provide good quality bug reports.
  • Test daily builds:
    Try run daily builds, reporting any issues you find in Blender's bug tracker is a great start.
    It's worth checking the development release notes (see: the link to the current development version) and test newly added improvements. Try to stress test too, check on corner cases or use cases that might have been overlooked.
    Or if you're really keen, check the commit logs though this does give a lot of detailed information to check over.
  • Check new bug reports: (also called bug 'triage')
    Even testing to see if you can redo a report can help,
    see the 'Needs triage' and 'Incomplete' categories in Blender's bug tracker.
    Sometimes developers don't have access to the systems needed to redo a bug, just confirming a bug or providing helpful details can contribute significantly to a fix.
  • Check new patches:
    You don't have to be a developer or write code to get Blender building from source, though you will need to learn a bit about some development tools.
    Once you have a build environment set up you can test patches which are still being reviewed. It's often useful for developers to get feedback on their patches from a users perspective.
  • Provide feedback:
    Once you have a login to developer.blender.org, you can comment on commits and design tasks. If you're experienced in a field and see room for improvement, feedback & insights can be helpful.
    Note, for this to work well there needs to be developers actively working on an area (or have time to investigate and respond to your feedback), Otherwise you may spend a lot of time to give feedback that can't be acted on, for this reason it's often best to focus on an area currently in development, more is written about giving feedback in other Q&A's here.
  • Improve Release Notes:
    Provide helpful example images for release notes, you will have to get a Wiki login first.
  • Improve/Write documentation:
    There are still many out-standing documentation TODO's, see the documentation project.
    This is an area that often lags behind, and we can definitely use help here.


This isn't a comprehensive list, there are other areas that can use help but these depend on more specific skill sets (improving/adding automated tests for example).

Why not have a place for users to make requests?

Q: I am interested to know what is the best way for community to voice their opinion on features or general development? (instead of littering the forums) For instance Autodesk has a prioritized feature list that is voted up by users which seems very efficient.

A: In principle, something like this could work for us (the topic has come up many times and it's been discussed amongst developers), we agree the ability to manage user feedback can work well. The issue is, running such a site successfully isn't 'free' - There has already been an effort to do this (http://www.blenderstorm.org (web archive)), which is an example of what happens when such a project is attempted without being moderated/managed.

Getting feedback is not typically a bottleneck for development, we have so much feedback already from existing communication channels, that employing a small group of people to manage it (moderate, reply, close-duplicate-suggestions, maintain website), takes time away from other areas. Some of this work definitely had to be done by experienced developers, who usually have deeper insight into why things are the way they are.

To put it differently: It would be great to have this, but we are concerned that we wouldn't be able to manage such a task successfully with the currently available resources.


See also: Best Place to put Feature Requests?

Update: Since this answer was written a new community run site has has started to handle feature requests rightclickselect.com, the site is currently in development.

How can users provide feedback to developers?

A: The user/developer ratio makes it difficult to respond to all feedback, however there are some ways you can communicate with developers.

  • Use the bug tracker if it's an error in the software.
  • Use the Blender Functionality (bf-funboard) mailing list for general user functionality discussion.
  • Use one of the Blender mailing lists which is focused on single topics (bf-animsys, bf-modeling, bf-cycles... etc).
  • Use the #blendercoders IRC chat room for more informal communication
    (though this can be hit-and-miss if you need to talk to an individual developer).

Developer Community

Whats the best way to become a Blender developer?

A: Unfortunately there isn't really a good one-size-fits-all answer to this. It depends so much on your own interests, motivation and experience level (also which areas of the software most need of attention).

Developers often give different answers, to this question when asked in IRC too.

So to stick to some basics:

  • Get Blender building from source.
  • Become familiar with a single area of the code you find interesting.
  • Start trying to make small changes & improvements.

For general info see: New Developer Info

Links to more opinionated advice (other devs feel free to link your own here)

Are there plans to improve docs for new developers?

Q: "I'm thinking of things like overviews/introductions of how different areas are structured etc (e.g. modifiers have the modifier implementation file + DNA modifer data etc, you need to make sure you specify how to write the data in the readfile.c /writefile.c files etc)."

A: Currently no short term plans, however we know our developer docs need work, we do keep basics up to date such as building, developer environments... but as for documenting big picture stuff, this is mostly older content that needs updating.

As for documenting how to add specific features, while we could do this, it's not clear if we have the time to maintain docs at this level. (It's been done in the past but the docs tend to get outdated).

You can always read previous commits of features that were added, but this isn't always ideal since API's may have changed since.

Another possibility is to have 'template-patches', which we provide that add common functionality.
(unofficial project, started but didn't get so far https://github.com/ideasman42/blender_patch_templates )

--Ideasman42 10:09, 6 May 2015 (UTC)

Organization

Is there a list of currently employed Blender developers?

A: not exactly, there isn't a full list of all the currently employed devs.

However reading over the developer fund you can get an idea of who's working on Blender:
https://www.blender.org/foundation/development-fund/

Also, developers who do work of Blender maintain a weekly log of their work.

Why can't we donate money for specific features?

A: While individual developers have accepted donations/payment for their developments before, this is not something any organizations have been interested to do this so far.

The main reason is that it's quite risky and complicated to develop very large features (the kinds that users may be interested to fund) with a high enough chance of success.

Sometimes projects are completed in a branch but eventually abandoned because the methods used don't work well and simply refining the approach won't give us something usable. If such projects are part feature-specific funding, this puts us in an awkward situation.

Along with this risk, it means developers end up being locked into a plan which we can't modify without possibly upsetting some of the supporters. (even when it makes sense to revise plans).

All that said, feature specific crowd funding has been shown to work, and individual developers are free to use online services to accept donations.

Note: pros and cons of different funding models is a big topic, this answer is intentionally a short summary.

--Ideasman42 17:26, 6 May 2015 (UTC)

Why is patch review taken so serious for a community driven, open project?

Q: The barrier for getting into Blender dev is already really high, why make this even worse by requiring intense review even though it should be possible for everyone to participate in such an open project?

A: While we're all interested in having fun in the development of the Blender project, we also want to provide a product with a high quality. Carefully reviewing submitted patches is one of the best approaches to archive this. Actually, the review progress in Blender can be split up into 3 different parts:

  • Feature Review: This means reviewing the feature itself. Is this a feature that fits into Blender at all? Is there a tangible and better way to reach the goal of the feature? Does the feature fit into the current roadmap (it doesn't make much sense to add features to parts of Blender that are planned to be rewritten). The decision over this is up to the module owner of this area
  • UI review: Ideally, the UI-team would review the UI of each patch before any in-depth code-review happens. However, since the UI-team has limited recourses, this is currently more a part of the general code-review. Nevertheless, checking with them is welcome!
  • Code Review: Usually the most important part of the progress. For a project of the size of Blender, it is *crucial* to ensure added code is good (More info about what good code means and why it is so important, here).

No doubt, this can be a tedious progress, especially for newer devs it is an additional hurdle and for reviewers it is additional work, but it also has been proven to help. If we really want to help prospective developers to get started, we should improve our docs, make code more readable and improve communication, instead of lowering quality standards needed to get a change accepted.

Design Choices

Feature X in Application Y is Perfect! Why can't you add it without
"re-inventing the wheel"?

Short Answer: Because Blender is a different application!

Long Answer: Blender has its own work-flows, follows different paradigms.

For any functionality of course we need to consider the options available - including any existing conventions.
However we also need to make sure this fits well with Blender which has its own conventions. It's important to think for ourselves, questioning the status quo since it may not be a good option for us.

We can aim to do better too!

This is not to suggest when Blender is different it's always 'better' or that developers always make the best design, only giving some rationale why simply cloning behavior from other applications isn't something you should expect developers to do.

This topic is subjective, user experiences vary widely!

Why not use an existing GUI toolkit?

Q: There are cross platform GUI-toolkits such as Qt, GTK, wxWidgets, Fox, FLTK, TK ... Why not use one of these?

A: The reason is both historic and practical: to move over to a widget set at this point would be a huge project and it's not even clear that doing so would be better for us. With complete control over our own toolkit, we have been able to support features not existing in other toolkits, (window splitting scheme, grabbing the mouse while operating sliders, widget control via the 3D-mouse, executing commands on widget mouse hover, multi-number button editing, our own data API's integration with the interface, browsing inside .blend files with the file browser, ...).

Another aspect to this is that we don't depend heavily on the GUI-Toolkit's release schedule. We can fix bugs in the interface code without depending on a third party toolkit team and its release plans.
Since Blender was written, widget-toolkits have come and gone.

Other advantages of using OpenGL include remaining lightweight, portable, uniform GUI across all platforms with the flexibility to change the GUI behavior as needed.

Of course this adds a maintenance burden, when we want to implement some features existing in other UI's - we need to do it ourselves.

There are pros & cons obviously it's possible to create 3D software without writing your own graphical-interface system, however at this point switching to a GUI-toolkit is a big project, so there would have to be a very compelling reason to do so.

Why are patches/features rejected just because "code isn't good enough"?

Q: A great patch/feature was rejected because "code wasn't good enough" or because it "doesn't fit into some code design", even though many users confirmed that it works nicely. Why is that? What does that mean exactly?

A: Unfortunately, it happens more and more that patches are rejected (or large changes are requested) because of issues with code, even if they seem to work exactly as wanted.

'Issues with code' or even 'bad code' doesn't mean that the code doesn't look nice because of formatting/code style, that the code is in the wrong place or that we don't like the feature itself.
In most cases, code that isn't good enough to be accepted is potentially dangerous. That means, it may break existing features, it contains known but hidden issues, or it makes future development difficult.
In fact, all three reasons happen more often than people think, even after extensive user testing. Further, the latest one is probably the most important and most common one. In the past, much of Blender's code was written without keeping possible future requirements in mind. Today, Blender's development suffers a lot from such 'bad' code (that's why so many rewrites/refactors need to be done lately). One can write code that works, but that doesn't mean that it will also work for the future.
Code may also be rejected if it's known that there are better ways (faster algorithms, better papers, more clean methods, etc) to get a task done.

Of course, exact reasons vary widely for each case.

For all sides, users, patch authors *and* reviewers, this is an at least unpleasant situation. If this happens, we try to communicate exact reasons for rejecting well. Finding a proper solution is in the interest of everyone, so we try to help with that, too.

Recent Examples: Fracture modifier (D817), Cycles adaptive sampling (D808), Cycles shadow-catcher, Alembic (rejected by Author himself), etc.

--Severin 15:56, 20 August 2015 (UTC)

Why not support languages besides Python?

A: Currently Python is serving our needs quite well as a scripting language.

Python has its limitations, other language/runtime combinations undoubtedly have their strengths. But Blender is primarily for artists - not an SDK/API to support developers preferred technologies.

Another factor is the effort to support/maintain embedding multiple languages (supporting Python is already giving us enough work / bugs to fix).

That said, the option is always open to embed other language/runtimes, for now though it doesn't seem worthwhile.

See more in-depth discussion here:

「https://wiki.blender.jp/index.php?title=Dev:Doc/AskUsAnything&oldid=149679」から取得