HNHacker News
TopNewBestAskShowJobs

rleigh

2,038 karma · joined June 15, 2015

Embedded software developer for medial diagnostic instrumentation
submissionscomments
rleigh··on We are shutting down Ondsel
It's because these tools are used to deliver high quality physical products on time and on budget with no scope or interest in messing around with software defects. These tools are treated like physical tools. It's worth paying for a high quality one which won't cause unexpected grief, by people who will provide immediate support because we pay them.

I think if there were good quality open source equivalents they would be considered, but they pose a huge risk, possibly even an existential risk, if they derail our development plans unexpectedly. Paying a lot of money for seriously good quality tools reduces that risk dramatically.

I've had a brief look at FreeCAD, and it's got a lot of potential. But when you compare it with SolidWorks, OnShape or SolidEdge, there's clearly a huge gap in usability and capability which needs closing before a lot of people will be able to consider it seriously. I'm sure it will eventually get there, like KiCAD did, but it will take many years and a lot of investment to get the usability, polish and featureset up to parity. It looks like Ondsel did a really good job to make some progress along that path.

rleigh··on The Netherlands has returned some stolen artifacts to Indonesia
They are both part of the United Kingdom (which is the union of the kingdoms of England and Scotland).

Great Britain the name of the island both are located upon. England and Scotland will always be a part of Britain, because that's a geographical area, not a political one.

rleigh··on Libadwaita: Splitting GTK and Design Language
You're trying to argue that water is not wet. This is seriously bad-faith argumentation.

Having a compatibility shim in GTK and testing that, once, would save any downstream user from having to do any work at all for this change. This is obvious and self-evident. Yet you seem to think that it's acceptable to impose this upon every user. It's this disrespect for end-users' time and resources which lead to people such as myself abandoning GTK entirely, when it was abundantly clear that it would not, and could not, ever be fit for purpose when you have this type of attitude prevail. Professional library maintainers do not break backward compatibility in such a cavalier manner, particularly when there is zero material net benefit of any sort arising from the change. Why would a project deliberately cause such breakage when it didn't have to. It's because it cared more about making a "cleanup" than it did about breaking it's entire userbase. It's cosmetic at best, and it could have been implemented without any break at all with a bare minimum of effort. That's the real kicker. The change could have been made without any compatibility break. And that just shows a complete lack of care.

Bear in mind also that Gtk*Box are foundational container widgets. Every application of any serious size will likely use hundreds or thousands of them. And no upgrade path for Glade/GtkBuilder XML either. That all needs hand-updating too. And this is just one example of breakage. You have to multiply it by all of the others, too. The ongoing burden of unnecessary and unproductive work repairing breakage arising from API churn is extraordinarily costly. Plus, it also breaks compatibiity of our application code with older GTK versions, which we might well also need to support in parallel for years. None of this adds value to our application, it's all cost.

You've spent pretty much all of your comments here deflecting and prevaricating. You've not once shown any concern of any sort for the actual real-world problems which have been imposed upon others, and which are genuine deal-breakers for actual application developers who have tried to use GTK for serious commercial work. The exact same lack of concern and understanding which the GTK and GNOME developers have shown all along. And I'm not new to this. I've used GNOME since pre-1.0 and developed with GTK+ since the 1.x and 2.x days. I was using GTK+ for commercial products two decades ago. It was barely viable then, and it's many times worse now. The primary concerns of these libraries should be API stability and implementation quality, and they have repeatedly failed at both.

If GTK wants to be considered seriously, it needs to behave seriously. And you need to actually listen and understand what people are telling you.

rleigh··on Libadwaita: Splitting GTK and Design Language
Yes, you could.

But this needs doing in every single downstream user of GTK. Far more efficient to do it once, in the toolkit itself.

This is why sane people don't use GTK. The maintainers literally couldn't care less that their 2 hours of "work" doing the removal causes hundreds of thousands of man-hours of refactoring on the part of others, plus the testing and validation work to prove that every part of their applications are still working correctly.

rleigh··on Libadwaita: Splitting GTK and Design Language
Yeah. It really does take more effort to leave perfectly working functionality in place than to do the work to remove it.

Come on. Many of the decisions made by the GTK+ developers are utterly unjustifiable.

Take the removal of GtkHBox and GtkVBox in place of GtkBox. A couple of shims for backward compatibility would have taken just a few lines of code and would have avoided a hard compatibility break. Did that happen? No. So a huge burden to update every GTK+ application (incompatibly!) was imposed upon every developer on the planet. Was that a productive use of resources? No. It was an utterly illogical change which had zero real-world benefit to anyone. And that's just one of many, many bad decisions.

There are very valid complaints to be made about GTK+, and you can't handwave them all away with some PR comments.

rleigh··on Libadwaita: Splitting GTK and Design Language
No. Why should anyone avoid discussing what happened?

We can all see with our own eyes how much GNOME cares about collaboration and interoperability with others. It's zero. It's been this way for a very, very long time. And that disdain for everyone else has consequences.

I used to develop GTK+ applications. I no longer do. Because it was an absolutely miserable experience, working with a toolkit which repeatedly requires every application developer to down tools and do a lot of busy work rewriting perfectly working code when APIs are changed or deleted. No other GUI toolkit causes so much pain and disruption to their userbase. It's quite clear that there is no regard for the actual needs of real application developers, and people like yourself aren't helping. You can't defend the indefensible.

rleigh··on Autoconf and configure features that people find valuable
They don't really define "complexity" well enough. CMake is more complex internally than the autotools in terms of the functionality available and the internal implementation details. But it does much, much more than the Autotools ever attempted. However, the user configuration is simpler and more powerful, and the generated build logic is simpler and more flexible in terms of the output formats available. And it only requires learning a single language, not three or four.

You're absolutely spot on with regard to its maintainability though. The Autoconf macros can't be changed without breaking everyone, so it's effectively impossible to change it, and we've seen that stagnation for well over a decade at this point. The use of m4 adds a whole heap of incidental complexity and inflexibility. You can never replace m4, because of the huge number of Autoconf functions internally-expanding user-provided macros.

I'm unsure why they make such a big deal of configurable install prefixes and system feature testing. While Autoconf may have pioneered this, or at least popularised it, it's not like the alternative systems like CMake don't implement the exact same features. They do, though with slightly changed names (just cosmetic).

Where I think they have a point, is in the huge investment into the Autotools by projects and the inertia that creates to resist change. That's very real. However, I do think that the cost of change is often overestimated. Off the top of my head, I've converted four open-source Autotools-based projects to use CMake (schroot, libTIFF, Apache Xerces-C++ and Apache Xalan-C++). Some of these had very complex logic encoded in their configure scripts; some of them had libraries of Autoconf macros to do custom feature testing. And yet, while rewriting this logic to use the CMake equivalents was tedious, it was entirely doable, and in all cases it was doable in under a week with a bit of shakedown afterwards to eliminate subtle behaviour differences. For two of these (libTIFF and Xerces), the projects maintain both systems in parallel to this day.

The inertia will keep a lot of established projects using the Autotools for several more years I would expect. If it works, the incentive to change is not going to be there until the Autotools bitrot to the extent that they break with contemporary toolchains. It will happen at some point. We already see it occasionally with libtool, and I've already seen the Autotools builds I maintain fail on Solaris, Cygwin and MinGW where they worked without any special effort with CMake.

Every piece of software is a child of its time. The Autotools were created to solve Unix portability problems of the '90s, and they served that purpose well using the portable and commonly-available tools of that time. However, while these tools continue to work today, they have not evolved to solve contemporary problems and that is why they aren't going to be a good choice for new projects. It will die by attrition if nothing else.

rleigh··on New Linux glibc flaw lets attackers get root on major distros
While using a setuid binary to edit the password/group "databases" is the historical default, there's no real technical reason why it must be that way. The passwd program could communicate with the database service via a socket. Likewise the NSS and PAM stuff could communicate with the same service via a socket. No reason for it to be lots of in-process loadable modules in this day and age.
rleigh··on Confocal microscopy
Confocal and multiphoton microscopy were the staple of my Masters and PhD research. I was hooked from the first time I got to operate one, the sheer beauty of what you can capture in true 3D volumes is on another level from conventional photography and brightfield microscopy. Multiphoton lets you live image cells and tissues and directly visualise the dynamics of how they move about and change over time, which gives you even more insight into the function of biological systems. Absolutely incredible technologies.
rleigh··on Star neuroscientist may have manipulated data to support a major stroke trial
I think the cost and complexity of reproducing work is somewhat overestimated, as is the specific expertise of individual researchers, though maybe your field is exceptional in this regard.

Primary research, pioneering new techniques and equipment to explore the unknown, is time-consuming and costly and requires a lot of original thought and repeated failure until success is achieved. However, reproducing that work doesn't involve much of this. It's taking the developed methodology and repeating the original work. That may well involve expensive equipment and materials, and developing the technical expertise to use them, but that does not involve doing everything from scratch and should not take anything like as long or cost as much.

I also believe that we far too readily overestimate the specific special skills which PhD students and postdoctoral researchers possess. Their knowledge and skills could likely be transferred to others in fairly short order. This is done in industry routinely. A PhD student is learning to research from scratch; very little of their expertise will actually be unique, and the small bit that is unique is unlikely to be difficult for others to pick up. I know we don't like to think of researchers as replaceable cogs, but for the most part they are.

My background is life sciences, and some papers comprise years of work, particularly those involving clinical studies. However, the vast majority of research techniques are shared between labs, and most analytical equipment is off the shelf from vendors, even the very expensive stuff. Custom fabrication is common--we had our own workshop for custom mechanical and electronic parts--but most of that could have been handled by any contract fabricator given the drawings. And the really expensive equipment is often a shared departmental or institutional resource. Most of the work undertaken by most of the biological and medical research labs worldwide could be easily replicated by one of the others given the resources.

Depending upon the specific field, there are contract research organisations worldwide which could pick up a lot of this type of work. For life sciences, there are hundreds of CROs which could do this.

As one small bit of perspective. In my lab a PhD student worked on a problem (without success) for over a year. We gave it to a CRO and they had it done in a week. For less than £1000. The world is full of specialists who are extremely competent at doing work for other people, and they are often far more technically competent and efficient than academic researchers.

rleigh··on FreeBSD 14.0 has reached – RELEASE
1. I use it on a NAS system. For years this was vanilla FreeBSD from 10 to 13. A few months back I replaced the system with TrueNAS Core which is based on FreeBSD 13, retaining the ZFS pools from the original installation. This system hosts storage and network shares, services hosted in jails such as databases, build slaves and artefact storage, and Windows virtual machines also hosting services and remote desktops.

2. First-class ZFS support, full NFSv4 ACL support which works with Samba and Windows ACLs and is a massive improvement upon POSIX.1e DRAFT ACLs.

3. No real preference, but ZFS support is (for me) the killer feature. Next are jails and Bhyve.

4. The main lack is the more comprehensive selection of drivers found on Linux. That said, it's pretty decent and I do find the overall quality of the drivers and system as a whole is better than Linux.

5. The system is engineered as a cohesive whole. While other BSDs might be similar, and perhaps Debian was attempting this a couple of decades back with its core design principles, most of the alternatives are lacking in this essential cohesiveness.

rleigh··on OpenZFS – add disks to existing RAIDZ
You can do a zfs send/zfs recv to send a dataset (including all snapshots) to yourself which is effectively rewriting the whole dataset, including all of its history, by duplicating it.

Not hard, but it does require sufficient free space. Once it's done you can destroy the original dataset and reclaim the space.

rleigh··on Gnu/Hurd strikes back: How to use the legendary OS in a (somewhat) practical way
QNX managed it! Drivers run in separate address spaces as, does the kernel, making it very robust. It's used for safety-critical applications.
rleigh··on Ask HN: Do we still need SWAP partition?
On ZFS, use a zvol as a swap partition. Easy to create, destroy, resize, have redundant copies for robustness, all of the configuration options you have for a data volume but use for swap.

I've done this on both FreeBSD and Linux, never had a problem. However, IIRC there might have been some issues in the past on Linux.

rleigh··on Don't Use Mocks
fff - https://github.com/meekrosoft/fff
rleigh··on Ask HN: Can Lua be remade to use 0-based numbering?
So in Lua the "first" element has an index of 1. There's absolutely nothing wrong with this. And most people who haven't already been exposed to C and related languages would find this perfectly natural. Because it's an index number, not an offset.

It's only in C and C++ where the memory location of an element can be expressed as base+offset that we end up using the offset rather than an index. That has clear logic behind it and absent optimising compilers it may be marginally more efficient, but it does not make for the most user-friendly interface--it does break the expectations of most other people.

If I'm working on providing a Lua library interface for non-developers, they will find the default Lua behaviour far more natural than were I to force zero-based indexing upon them.

rleigh··on Trying Gnus as an Email Client
Yes, it was the newsrc.eld which got corrupted. I can't remember the circumstances, but I think it was an Emacs or Gnus bug at that time. This would be over 10 years ago though.
rleigh··on Ask HN: Anyone else feel like their whole career will just be tech debt?
This is one of the reasons I moved into embedded software. It's much more satisfying for me personally. But even when I worked on business software, it was nice once in a while to go out an meet the actual customers who used the stuff day-in day-out. I learned a lot from them, and took it back to improve things, and that made it quite a bit more interesting, knowing I was addressing real usability and functionality problems which would improve their job.
rleigh··on Ask HN: Anyone else feel like their whole career will just be tech debt?
I have seen this repeatedly over the years. It's frustrated me personally many times.

However, just to offer a counterpoint, one thing I have seen time and time again from highly-motivated technical people is that they place the new and exciting technologies above the actual business needs. There is an inherent risk in using cutting-edge tools and technologies, and if the business is working well with the current stuff, why take a risk you don't have to?

Of course, the other end of the spectrum is only using completely obsolescent tools, and that can be a problem in and of itself as well, because it also brings risks with it.

I have, unfortunately, been on teams where these people got their way and it led to very expensive project failures because when reality came to bite, they couldn't deliver. The new stuff couldn't meet the requirements. Using existing technologies which were known quantities would have avoided that risk entirely, even if it wouldn't have been as fulfilling for them personally. The company isn't run for individuals' personal amusement and tech fetishes, and sadly when that takes priority over all else, it can lead to failure on colossal scale.

I still try out new stuff, but I don't propose we use it for serious business-related activities until it's been fully proven out.

rleigh··on Trying Gnus as an Email Client
I used it for years until one day it corrupted its own cache and become thoroughly unusable. Manually repairing tens of megabytes of serialised Lisp isn't fun or practical. If you live in Emacs, it's a great tool though.

Sadly, I had to move on to Thunderbird, and lately Outlook. Sad to say it, but email has become an absolute pain to use and text-based email clients are increasingly unusable. At some level, I feel I'm a sell-out. I no longer self-host my own email server, and I no longer use Gnus or mutt. It does feel like the rest of the world has forced us off these high-quality clients as it has consolidated around a small number of providers who deliberately prevent us doing it ourselves or using our own clients without jumping through major hoops like authentication, and this is a big loss.

rleigh··on Create texinfo files with org-mode for your personal notes
While I can understand using org-mode, why on earth would anyone use Texinfo in 2023? It's so thoroughly obsolete, and completely eclipsed by HTML. The "TeX" part is also thoroughly limited in its capabilities. It's always been a somewhat odd format, being both more structured than troff/nroff, but at the same time being vastly more limited in features, and is something of a subset of both TeX and HTML while excelling at neither. It's table and figure support is not at all great, and its graphics capabilities are very limited. If you absolutely have to live within Emacs, then maybe "because it's there" is sufficient justification, but I doubt this applies to most people.
rleigh··on New study: 70% of type 2 diabetes cases linked to food choices
Well, given that he references papers from the "corn refiner's association" immediately afterwards, I'll note that this may not be entirely unbiased. This entire industry in the US has immense political clout, and that may well extend to publishing questionable research that comes out in their favour to confuse wider debate and regulatory change.

There's plenty of evidence from the other point of view demonstrating that HFCS has rather different behaviour than sucrose. Quickly-found example: https://pubmed.ncbi.nlm.nih.gov/22152650/

rleigh··on New study: 70% of type 2 diabetes cases linked to food choices
There's a very big difference between HFCS and sucrose. HFCS is soluble monosaccarides. Sucrose is a disaccaride which requires enzymatic cleaving to produce free fructose and glucose. The kinetics of the fructose processing are very different as a result--with sucrose the release is much slower because it's rate-limited by the quantity and kinetics of the sucrase enzyme. It's not going to hit your liver with a high concentration of fructose instantaneously, it's going to be a slower release over an extended period.
rleigh··on New study: 70% of type 2 diabetes cases linked to food choices
However, glucose can be stored and metabolised throughout the body whilst fructose requires processing in the liver. Our bodies are not designed to have a large instantaneous "hit" of fructose. They can cope with slow release of natural fructose in fruits, but in processed food it's soluble and released very quickly. The kinetics of how it's processed are very different.
rleigh··on America’s first high-volume ‘PFAS Annihilator’ is up and running in W. Michigan
What was wrong with the old method of using waxed paper? Non-toxic and waterproof. They did of course eventually go soggy at the ends but it's somewhat incredible that we go for known toxic coatings instead of previously-used non-toxic alternatives.
rleigh··on Current issues with the Qt project from the outside looking in
This is exactly the reason I've avoided QML. Give me a 100% C++ API and I will reconsider.
rleigh··on The GTK+3 port of GIMP is officially finished
Before 2.0 it was GtkObject. 2.0 just factored it out so you could use it for non-GUI code as well.
rleigh··on The GTK+3 port of GIMP is officially finished
The XML isn't upgraded and there are no tools or XSLT to do it. You end up hand-editing the ui files by hand. Not fun or productive.

There are good reasons why professional developers abandoned GTK+, and this is just one of them.

rleigh··on Unpopular Opinion: Don’t Use a Raspberry Pi for That
I think you're pretty much correct. Whether these details matter is entirely application-specific, but you can go the extra mile if your application requirements demand it.

There are certainly multi-core MPUs and MCUs with a mixture of cores. The i.MX series from NXP have multi-core A7s with an M4 core for realtime use. Some of the ST H7 MCUs have dual M7 and M4 cores for partitioning tasks. There are plenty of others as well, these are the ones I've used in the past and present.

rleigh··on Unpopular Opinion: Don’t Use a Raspberry Pi for That
As I mentioned in the previous reply, the CPU caches and the MMU to begin with. You're probably running your application from SDRAM and an SD card. The caches and page tables result in nondeterminism, because the timing depends upon existing cache state, and how long it takes to do a lookup in the page tables. And as soon as you have multiple cores, the cache coherency requirements can cause further subtle effects. This is why MCUs run from SRAM and internal FLASH with XIP, and use an MPU. It can give you cycle-accurate determinism.

The A-profile cores are for throughput and speed, not for accurate or consistent timing. However, you can disable both the cache and the MMU, if you want to, which will get you much closer to the behaviour of a typical M-profile core, modulo the use of SRAM and XIP. If you're running bare metal with your own interrupt handlers, you should get good results, excepting for the above caveats, but I don't think you'll be able to get as accurate and consistent results as you would be above to achieve with an MCU. But I would love to be proven wrong.

While most of my bare metal experience has been with ST and Nordic parts, I've recently started playing around with a Zynq 7000 FPGA which contains two A9 A-profile cores and GIC. It's a bit more specialised than the MPU since you need to define the AXI buses and peripherals in the FPGA fabric yourself, but it has the same type of interrupt controller and MMU. It will be interesting to profile it and see how comparable it is to the RPi in practice.

← PreviousPage 2 of 33Next →