User:HimanshiKalra/log

Inspired by Hans Goudey and in case anyone is interested, I would like to document my remaining summer here, although already more than half of it is over, and I will try my best to be consistent and not turn it into my personal diary ;)

---

July 24

---

Today I worked on testing the Particle system, and finally tested the Fluid modifier, there have been some significant changes from 2.83 to 2.90, the test was failing because I was testing with different versions. The Particle system's Render property needs to be set to Object to be able to convert it to Mesh and be testable, I am not yet fully able to confirm whether it is reproducible or not.

July 25

---

Today, it was weekend time, didn't work on anything blender related, participated in a hackathon, which went fine. Solved one problem out of two, attempted all MCQs, mostly cyber security related questions about BGP(Border Gateway Protocol) and something called YARA rules. Was trying to get Windows updated to get WSL-2 working on the system, so that I can run Docker, so that I can build my resume using Latex, and I don't know how to use any of that :D, the most long way to do it. I have nothing to write in it anyways.

July 26

---

Today, I was exploring WSL. Got WSL-2 running, added Ubuntu as a distro, I would have preferred CentOS was looking forward to check it out, but it is paid :(, got Docker running as well, exploring how to use it. Started with one of those projects which I rarely finish. The main reason for setting up WSL was to have a Linux system to run unit-tests for Blender in a different environment, and get faster builds. So it is partly blender-related. Got some basics cleared up for the Docker, won't really need them anytime soon though. End of weekend.

July 28

---

Today, I worked on testing Particle systems, they are reproducible after we set the seed and keep the world coordinates same. There is a kind of inconsistency in the simulation of Particles, say when you add the Particle system modifier at some frame n, the simulation is different if you add it to the first frame and "play" and "pause" it at frame n. Maybe "physics" is being re-calculated, should open a devtalk thread for this.

July 29

---

Yesterday I tested the Particle System, it was reproducible using a python script but hard to reproduce manually, it kind of took a lot of time, I wasn't sure why I can't reproduce it, so I was trying out different stuff, maybe world coordinates affect it or maybe setting a seed value will make it reproducible, all of it was just too fuzzy and not systematic, I reached no conclusion apart from "Just use the damn script!" I added a separate class for it and submitted a diff, next plan is to add Dynamic Paint modifier.

July 30

---

Today, I worked on Dynamic Paint, I thought I was half-way through with it in the afternoon, but now I just can't access the canvas settings from the framework, its probably my fault somehow in the implementation but I used the same technique for Fluid and Cloth and they are working fine, they are not complaining at all. Will have a look again tomorrow and would narrow it down with a script.

July 31

---

Yesterday, I had a brief meeting with Habib to start the refactoring process, the approach discussed to parse the settings is using a Tree based Depth First Search approach. I have had implemented that as a part of my curriculum in C++, using nodes and pointers, have no idea how to do that in Python. I will think of something. The Dynamic Paint Class functionality was successfully merged with Fluid Class, both are testable.

Aug 4

---

Aug 1 and 2 were weekends, Aug 3 was festival day, I was busy. So yesterday, I worked on traversing a generic tree(not binary tree) in a pre-order way, given that the tree is already given, I tried to use the similar approach to traverse a nested dictionary to be able to set nested settings of modifiers.

Aug 5

---

Today I implemented the traversal of nested parameters within the framework, its up for review, the code looks a bit "fishy" but it works, its recursive and has nested try-catch, I will try to simplify it further

Aug 7

---

I missed to write on Aug 6 :(, I am writing this on Aug 8, so I use the word "yesterday". Aug 6 - I worked on simplifying the code and made the changes as requested in the review, converted the nested try-catch to an if-else. Yesterday, I worked on merging Dynamic Paint and Physics Spec into Modifier Spec, I added frame support, updated the "_bake_current_simulation" method to merge different types of baking under one function only. Oh and there was another meeting with Habib to discuss "Possible next steps" to be done for the remaining 2 weeks.

Aug 8

---

Yesterday I added a test for Particle Instance and created a final diff for all the work I have done, its less compared to other students. Added Bastien as a reviewer to my diff D8507, tried to switching to my svn soc branch I created, should have logged that...The files started re-downloading, it is always a blind error, I am in the wrong path most of the times that's why.

In the end, I think Fluid simulation has randomness and in it and is not reproducible, even increasing the threshold doesn't help, so I ended up removing the Test for Fluid, as it was failing now.

Aug 9

---

Committed to my svn soc branch, added new and updated blend files, updated the Tools/Tests/Python page for Windows users, plan to add more tests and add documentation for "How to add more tests" and run them.

Okay here is the trick for svn, once and for all, for future me, while switching branches using svn switch we should be in the same path as relative link, in my case if I am in lib\tests, the svn command would be svn switch https://svn.blender.org/svnroot/bf-blender/branches/soc-2020-testing-frameworks/lib/tests. It is different from checkout in which we should be one step outside ../ and then checkout.

Aug 10

---

Today I worked on adding more tests for Cloth modifier, added tests for spring, self-collision, updated the documentation for testing including the "why automated testing is needed" and "how automation" works. So the newbies don't end up doing it manually like me. I thought I would be busy today, not so much.

Aug 12

---

Less code work was done on Aug 11, I was looking at Release notes and bug tracker, updated diff to include new cloth tests, merged master locally and pushed to remote. Today I have been making improvements as per the review.

Aug 16

---

I took off for around 4 days (Thursday to Sunday), I had a coding round, did a little preparation, didn't go that well. Anyway, now its over, so I am back to Blender.

Aug 17

Today I am working on my test object generator script, I am trying to abstract it so that the user can interact with it better, without being exposed to the code much. I haven't worked on it for quite a while, so its a bit rusty. My college classes are supposed to begin tomorrow.

Aug 18

Today I worked on extending the generator script, basically adding another class to handle the "file handling" part, generation and creation. There is no fully correct way to check for path validity or I am yet to find it out, I did it the old way using os.path.exists() and asking for the filename as well as path, there are 2 approaches either I can do it in the code, somehow split the string and extract the filename, and path separately to check for path validity or just take it from the user, I am going for the latter approach, its easier and the user just have to type it once anyway. The new GenObject class will take a list of TestObj. I managed to make it Exception safe, proper unlinking from Scene/master collection is left.

Aug 19

---

I am a little blank on what I worked on today, its vague, some testing, some work on script, and going through the review

Aug 20

---

Today I worked on improvements suggested by Bastien in the review, code-formatting, renamed "test names" of bevel and boolean operator tests, and I accidentally broke the operator tests, the verdict for Cloth Spring and Self-Collision is non-reproducible, I will test again after compiling on Ubuntu, finally WSL will pay off, already cloned it, the libs are left.

Aug 21

---

Today I worked on the generator script finally its up for review, there is a scope of adding back the Vertex group functionality, I will look into it later. Finally compiled blender on Ubuntu. I also got feedback on User Documentation wiki page, so I worked on it and updated it. I also started with the documentation of the Final Report to submit it to Google as the Work Project.

Aug 23

---

I tried getting some review work done on Saturday but got bored mid-way, so took the weekend off. Sunday, asked Bastien for a meeting and updated the weekly report.

Aug 24

Today I am working on creating a new common class for ModifierTest and DeformModifierTest from which they both can derive, its called RunTest for now, I am facing some implementation problems...

Aug 25

The original idea was different, it was to inherit from a parent class and one of the parameters of the method of the Parent class was a child object, hence it was a bit crazy but it worked.

The current implementation involves using an Abstract Base Class (ABC) and to have a check for run_test for the different Helper Modifier Test classes, this is what I worked on today. There is some discussion on merging both the Helper classes.

Aug 26

Removed the abstract base class, as there is now only one Helper class, merged DeformModifierTest and ModifierTest, so now there is only ModifierTest class. Today I worked on Final Report, it took quite some time to gather all the work I did and put it in words.

Aug 27

Today I worked on removing Operator Test class and introduced a generic Run Test class it can be used by both Operators and Modifiers, revamped bevel, boolean and operator test files, used everything to minimize time, find and replace is THE BEST, then later I used "multiple cursors", it also increases speed. Overall it was tiring but its done, it is generalized which is a good thing.

Aug 28

Today I was working on my Btech project in the morning, I made some tweaks in the final report, there were some issues with my remote branch being diverged from my local branch. And I submitted the evaluation. I think I don't intend to work on the weekends, might give a look at Generator script. So its mostly a wrap-up for me.