Previous Up Next

Diary, April 2026


Sun Mon Tue Wed Thu Fri Sat
              1   2   3   4
  5   6   7   8   9  10  11
 12  13  14  15  16  17  18
 19  20  21  22  23  24  25
 26  27  28  29  30

Wednesday, April 1, 2026

Lone signpost in Mongolia

In episode 128 of season 8 of Itchy Boots titled 'Loneliness hits different in Mongolia 🇲🇳', Noraly Schoenmaker comes along a signpost in the middle of nowhere as can be seen starting at 13:53. This is along the grevel road A1701, which at many place is not a single road, but many parallel roads often spread out quite widely. So, it is rather interesting that there is a sign somewhere in the middle when most local just take one of the many tracks. I found a photo on Alamy.com that was taken by Daan Kloeg on Augustus 7, 2019. I contacted him and he promptly gave the location from which the photo was taken, allowing me to find the signpost on Google Earth. According to the satelite image on Google Earth the location of the signpost is 48°55'59.88"N 92° 2'19.38"E, a little west of the road on google maps and open street map, which might explain why I had not yet found it when I followed the road. (I am not sure whether I followed the complete road from the start to the end.) A little bit further (at 48°57'44.50"N 92° 2'34.98"E), she comes along another 'sign' with the words (if I have transcribed them correctly): 'өлгий' 'тавтай морил', and 'сум', which I understand means 'Welcome to Ölgii district. A little further she travels through a village called Ölgii.

Thursday, April 2, 2026

Infinite mazes

I have been thinking about the construction of infinite mazes. When I say maze, I mean a maze on a infinite square grid (in all directions) where there is some way to calculate if there is a wall on each line segement of the grid and that the mazes is 'prefect' in the sense that the mazes has no cycles and that each square can be traveled to from every other square. The page Maze Classification give some algorithms for creating finite mazes on a rectangle (or square) of a square grid. There are only two algoritms: Wilson's algorithm and Aldous-Broder algorithm. For animations of these algorithms see the page Maze Algorithms. There is a simple method for generating infinite mazes from finite mazes of some fixed size n. The idea is to place grids of size n side to side on the grid. Asuming that the walls on the outside are all closed, it does not meet the requirement of a perfect maze. The idea is to create a maze of size n×n with 'rooms' of size n and have the openings determine where to add an opening between the mazes of size n. This should be repeated for every power of n. Asuming that the generation starts at the origin, one only has to generate this for a limited number of powers, depending on how far one moves from the origin. This technique is also explained on the page Making an Infinite Maze. This method does not really result in random mazes, because these kind of mazes will contain very long walls that are getting longer and longer the further one gets from the origin, which is rather rare in truely random mazes. How can be improve on this assuming that we start generating the maze from the center and do want to store as less as possible information, potentially being able to regenerate parts of the maze on demand? In some game, for example, the user would only see a limited part of the maze from above, or till a wall (or a certain distance) when from the inside. One idea is to generate the maze in chunks of n by n squares. In such a chunk it should not always be the case that there should be a path between all the squares. The idea is to start with an infinite grid that has no walls and start randomly adding walls inside a certain chunk as long as each sqaure still has a path to the 'outside'. If an additional chunk needs to be added, it will also be allowed to put walls on the 'sides' where the chunks touches chunks that already have been filled. Although this seems a good approach, it seems that it will reduce a different maze depending on the order in which the chunks are added. Even if everytime the same random sequence is used for each chunk, there is still a problem for the walls that needed to be placed on where the chunks meet. A solution could be to always generate the chunks in a 'fixed' order. If the order is truely fixed, for example in s spiral order around the center, one would have to generate more and more chunks if one is moving away from the center. A solution to this to introduce a dependend relation to the grid of chunks, were a chunk can only be generated when all the chunks it depended on are already generated. Maybe there is a way to define the dependend relation (with all other than the origin chunk depending on the origin chunk directly or indirectly) such that only a limited number of chunks need to be generated, possibly linear to the 'distance' to the origin.

Monday, April 6, 2026

Unexpected error messages

Since I finished replacing stage0 of live-bootstrap for x86 (32 bits) two months ago, I have been working on the x86 64 bits version. I heve now come as far as being able to compile the Tiny C Compiler sources, but the resulting compiler when compiling some sources gives some unexpected error messages, which seem to be correct (see commit 96ad3732). This will require some clever debugging. Maybe, should first test this compiler on some other sources.

Tuesday, April 7, 2026

All flowers open

It looks like all flowers on our magnolia have opened. The first flowers opened on March 18. The leaves of most of those first flowers turned brown due to frost and have partly fell down already. The latest of the flowers that opened are on the side of the house, which receive far less sunshine than the side where the first flowers opened. We did get flowers for quite a long period this year. The flowers spread a lovely fragrance.

Tuesday, April 8, 2026

Exhibitions

This afternoon, I visited the exhibition Over Water with photographs by Joop Overkleeft. It has a series about soldiers placing a Bailey bridge and some from his digital graphics.

Next, I went to Concordia and watched the Idem et Idem exhibition again. One of the artists was present and I talked a little about her project. I also saw the video with the work of Kore Danaë Plunkett that that I cannot remember seeing the last time (probably because it was not projected). It was kind of a recording of a performance by her. The translation of the description of her work is: "My life is defined by duality and contradictions, as a result of my limitations. Chain armor and two-dimensional work are the spaces in which I explore these characteristic aspects of my identity."

In the evening, I went to the exhibition RE:-)CONNECT at B93 with works by second and third year students of the AKI. There were individual works by the following students on display:

They also had some collective works in coorporation with Tessa de Roo. There were two screen print collective works and two collages of works on the wall. I found those quite interesting. I talked with one of the students about this because you do not often see that artists worked together on an art work. I understood that it had come quite natural when they started working with screen printing.

Thursday, April 16, 2026

Rijksmuseum Twenthe

Today, Conny and I visited Rijksmuseum Twenthe. We first saw the exhibition Pekka Halonen. An Ode to Finland with paintings by Pekka Halonen. I found the following works noteworthy: Really a lot of very impressive paintings. After this we saw the exhibition: Edwina van Heek: Silent force behind Rijksmuseum Twenthe, which honours that a hunderd years ago Edwina van Heek-Burr Ewing signed the deed of donation by which she donated a new museum building and an accompanying art collection to the Dutch State. Some of the works on display:

Next we saw the exhibition Many Loving Arms with works by Bas Kosters. I did not have high expectations of this exhibition, but I was quite impressed by it. I found the following works noteworthy (if I have not mixed up the wall labels):

We saw the exhibition Calculating Empires again. Conny took the picture below of me paging through the book Calculating Empires. On the wall I found the words Incompleteness and Undecidable but I could not quickly find it in the book. I wondered whether it is possible to buy the book, but (at home) I could not find it.

Sunday, April 19, 2026

Infinite mazes

I have been thinking some more about generating infinite mazes since I wrote about it on the second of this month. If you always generate the chunks in the same order, it is sufficient to remember which walls are connected on the perimeter in order to avoid areas that cannot be connected and that there are no cycles. However, generating the chunks in the same order is not very efficient. Whenever two or more walls on the outside are connected, one has to decided which one will be going to infinity or in other words for each wall we need to indicate in which direction we have to walk, while touching the wall, to arrive at the other side of the wall. In the other direction this will never be the case. I think this will work, but I get the impression it will result in some bias. Because walls going outside of the border can never return into it. A way to counter this, is also to determine how walls outside the border have to be connected. In interesting question is how many of these should be connected to avoid another bias? I fear that some serious mathematics is required to establish this. Maybe is also depends on the shape of the perimeter.

Thursday, April 23, 2026

Still the same error message

I am still getting the same error message with the 64 bits version of my replacement stage0 of live-bootstrap when I am using the interpreter that I wrote for Stack-C, the intermediate language that the compiler uses. In the past weeks, I worked on getting the interpreter working for the 64 bits version of Stack-C and I also implemented a debugging mode allowing to set breakpoints and step through the execution. Getting it to work took more time than I had expected. Now that they show the same behaviour, it is safe to conclude that the bug is likely not in the programs used to generate an executable from the Stack-C file but in the compiler itself or how the compiler is called (with respect to environment variables). The message both versions return is: 
lib/linux/x86_64-mes-gcc/crt1.c:55: error: non-zero addend on REL architecture

Monday, April 28, 2026

Sci-Bot: hamiltonian cycles

Yesterday, I tried out Sci-Bot with the question: 'hamiltonian cycles.' I found the result quite interesting and decided to keep the results in a local copy. Because the saved file contained UTF-8 characters, I wrote the program UTF8toHTML.cpp to convert these characters to HTML escape sequences. It looks like the AI bot is no longer workable without creating an account.

Wednesday, April 29, 2026

Fixed bug in C compiler

I finally fixed the bug in tcc_cc.c, the C compiler in my replacement stage0 of live-bootstrap that I have been struggling with in the past weeks. It turned out that the bug was in the preprocessor that did not handle the following code correctly: 
#define VAR_A 1
#define VAR_X VAR_A
#if VAR_A == VAR_X
#endif
It is important to realize, that the following, should return the same result: 
#define VAR_X VAR_A
#define VAR_A 1
#if VAR_A == VAR_X
#endif
This shows that macro definitions are only expanded when they are used. This is the same for the expressions for the #if directive as it everywhere else. In the iterator solution I implemented, I decided to have the expansion iterator only work after the condition iterator (that evaluates the various #if directives), because otherwise it would also expand all branches that are not needed and it would have something to not expand the macro definitions. (Maybe there is an elagent way to implement it, but I have not looked into it yet.) This meant I would have to implement something for the macro symbols in the expression and assuming that the full expansion rules would not be needed. I simply added some code repeatedly expand macro definitions in the expressions with two lines of code. This again shows that the implementation of the C preprocessor is far from trivial. I am sure that my implementation in tcc_cc.c is not compliant with the standard. There is still some other bug to be solved as I am no getting error messages about 'Invalid relocation entry.'

This months interesting links

Home | March 2026 | May 2026