I am trying to specify a series of discontinuous indices for an array:
subtype PIN_SEL_Index is Positive with
Static_Predicate => PIN_SEL_Index in 0 .. 4 | 7 | 9 | 10;The compiler complains:
error: subtype "PINSEL_Register_Index" has predicate, not allowed as index subtype
Is there a reason why this isn't allowed? Is there an alternative?
7 replies
What? Ada has some trippy syntax and semantics at a spooky distance! It raises more questions. First, are the indices Bar ordinal, actual position in the array, or nominal, just a name but isn't the actual position (i.e. maps to the contiguous range)? Second, 'First_Valid...'Last_Valid is a surprise. How is it different from 'First..'Last? Finally, can Bar contain a static predicate?
1 reply
Wait a second. Ada overloads syntax keywords all over the place but for predicates, it introduces the attributes 'First_Valid...'Last_Valid instead of recycling 'First...'Last? Does the latter work with predicates also? If so, how are they different?
First and Last gives you the bounds of the type/subtype without taking into account the predicates of that type/subtype. First_Valid/Last_Valid takes into account the predicates of the type/subtype.
1 reply
For MCUs, where is the delay/delay until "operations" defined? They appear to be baked into Ada.
4 replies
OK. I thought it would be a nice to have in the LightRuntime.
@barryhouser:matrix.org . For MCU delay until is defined in the arm-eabi folder (if targetting ARM), see screenshot
If you change anything in those files you need to gprbuild the runtime again
@aiunderstand: that is the Ravenscar-full profile :) That is like a small RTOS on its own and tends to be a bit customized for each board. I suppose ARM is fairly portable in that regard.
Well, to be more precise, a date, a time, a date and time, and so on and so forth. I don't think a simple timer would suffice for that. But if you just wanted to delay and no delay until, you could throw program_error when a delay until statement is found and you could just rely on relative delays
"The Sociology of Programming Languages" by Leo Meyerovich, from Strange Loop
2 replies
Just an update, the alire installer I've been working on compiles and installs Alire on
- x86_64 and aarch64 MacOS
- x86_64 Linux (via WSL)
- x86_64 Windows (it just bails out, but it WILL compile lol)
I still need to test it on a couple other versions of linux + freebsd, but I should have something on github shortly for more people to test.
4 replies
:point_up: Edit: Just an update, the alire installer I've been working on compiles and installs Alire on
• x86_64 and aarch64 MacOS
• x86_64 Linux (via WSL)
• x86_64 Windows (it just bails out, but it WILL compile lol)
I still need to test it on a couple other versions of linux, but I should have something on github shortly for more people to test.
• x86_64 and aarch64 MacOS
• x86_64 Linux (via WSL)
• x86_64 Windows (it just bails out, but it WILL compile lol)
I still need to test it on a couple other versions of linux, but I should have something on github shortly for more people to test.
fyi, I created a new category on the Ada forum for Advent of Code https://forum.ada-lang.io/t/about-the-advent-of-code-category/112
So I've got a bit of a question involving subunits. If I have an overloaded subprogram, but I want all overloads of that subprogram to also be subnits, how does gnat handle it? Does it require that all the functions of same name but differing profiels be in one file, or do I have to do something special like put numbers in the file name or something?
The defining_identifiers of all body_stubs that appear immediately within a particular declarative_part shall be distinct.
So you can't two overloading subprograms to be subunits.
1 reply
See the example here: https://sourceforge.net/p/gnoga/code/ci/dev_2.1/tree/src/gnoga.gpr
package Naming
I already found a tutorial using alire and gnat_arm_elf, but I couldn't get it to work on our dev board (imxsx with a cortex a9).
is there any way without writing my own runtime? like a flag or something? :s
you can try to use the bare_runtime with
-march=armv7-a
if you don't want to port bb-runtimes, then your best option is to run Linux or an RTOS
it looks like i.MXS is a very old chip... I'm not sure it's worth the effort unless you have some existing hardware you absolutely cannot change
changing it is not really a possibility. our board is deployed a few 100000 times at the very least. We have older boards too, but right now i'm just checking if we can run ada code at all
@hardliner66, you presumably have a host environment, I’m guessing Linux, where you cross-build code to run under Linux on your target board. If you can run GNAT on your host, you’re I’d expect most of the way there.
You might find the basis of an approach in the first half of this script runs an x86_64 Ada-capable compiler (on an M1 Mac; not sure this is necessary) to build an x86_64 cross compiler to aarch64 (Apple silicon), which might be all you need; the second part uses that to build a native aarch64 compiler, which you wouldn’t need unless you want to compile on the target.
Of course, there’s a whole load of mac-isms in there!
You might find the basis of an approach in the first half of this script runs an x86_64 Ada-capable compiler (on an M1 Mac; not sure this is necessary) to build an x86_64 cross compiler to aarch64 (Apple silicon), which might be all you need; the second part uses that to build a native aarch64 compiler, which you wouldn’t need unless you want to compile on the target.
Of course, there’s a whole load of mac-isms in there!
I think I may have failed to mention a couple of points:
Macs (x86_64, aarch64) are very very similar; you’ll probably need to have a copy of the target libraries on the host system; the last time I did this was in 2005 or so, linux / vxworks 5.3
In all cases, building a cross-compiler is best if you have a host compiler built from the same sources.
Macs (x86_64, aarch64) are very very similar; you’ll probably need to have a copy of the target libraries on the host system; the last time I did this was in 2005 or so, linux / vxworks 5.3
In all cases, building a cross-compiler is best if you have a host compiler built from the same sources.
1 reply
I feel like we need an Ada compiler that's purely LLVM-based so we can avoid the msys2 weirdness on windows. Or try to. I just set up Alire using the installer and bash kept giving memory sigbase mismatches or something like that... The error was the familiar
0 [main] bash (2588) C:\Users\ethin\.cache\alire\msys64\usr\bin\bash.exe: *** fatal error - cygheap base mismatch detected - 0x1289408/0x11D9408. The fact that errors like this still happen is just ridiculous IMO.
The odd thing is that I don't really have multiple cygwin installs on here. I mean I have the coreutils package and related friends from Scoop.sh but still...
Ah, it seems to be ASLR related. Will go mess with that -- though why these binaries don't have ASLR enabled by default nowadays is beyond me
I've published the initial release of an installer for alire: https://github.com/AJ-Ianozi/getada
I try to be clear in the readme, and I put some more details here: https://www.reddit.com/r/ada/comments/yzx6hz/getada_an_unofficial_installer_for_alire/
Irvise: do you think AURA needs to be in pkgsrc? https://annexi-strayline.com/blog/posts/4 or gcc-aux is enough for now ?
3 replies
Advent of Code - an online programming challenge has started today. Join us with Ada (or even SPARK) solutions and have a fun! 🥳
1 reply
Doesn't mean you can do "unit & integration testing, examples, docs", but it will be different than how you do it with cargo.