scripts VS C++ Middleware Writer

I made this at one point to counter that anti-Linux FUD; I still maintain it but not as religiously -- it's become too polarised and no one on that sub is logical once they've made up their minds (which perhaps applies to me also, but shrug !)

try https://github.com/xkcd386at/scripts/blob/master/restic-find-fzf if this interests you. Like a lot of my tools, this one also uses the amazing fzf program to help you select files or directories to extract

for people who want this kind of combination but still have muscle memory for head and tail arguments, try this: https://github.com/xkcd386at/scripts/blob/master/ht

if someone wants a pure bash+fzf+ripgrep solution with slightly different behaviour, take a look at https://github.com/xkcd386at/scripts/blob/master/vgc

So... https://github.com/xkcd386at/scripts/blob/master/fclone (and yes the program name is probably even less imaginative than rfync -- suggestions welcome)

I have a script that identifies exact duplicate directories; see https://github.com/xkcd386at/scripts/blob/master/dirdups

https://github.com/xkcd386at/scripts/blob/master/usff for the win

as for session token stealing, on Linux at least I have long advocated using multiple userids to run browsers -- I log on to gmail in a user called x3-gm, to amazon in a user called x3-amz, and so on. All simultaneously and without having to logoff/logon. https://github.com/xkcd386at/scripts/blob/master/usff has the details, but that's not important; the main idea is to not use the default locations for any important logged in site. Sure there is some pain in manually moving files between userids (e.g., move statement from bank userid to gmail userid to send to my accountant), but it's not hard and some judicious use of /tmp or similar can help a lot

I use unique_ptr, but not as much as I used to. I've never used shared_ptr. This is my library that uses some C++ 2020 and 2017 features.

I started with Linux in the late 90s. I switched to FreeBSD around 2013 and returned to Linux a couple of years ago. Io_uring was the main reason I had to come back. At first I ported the back tier of my code generator back to Linux and then I ported the middle tier from being POSIX based to Linux.

Are there others that minimize multithreading and opt for multi-processing with single threaded processes? Call me a simpleton, but this approach eliminates some of the most difficult bugs by design. Here's an example of one of my single-threaded servers. The network io is asynchronous, but the file io is synchronous. Thanks

I've been working on this program for 13 years now. At one point it had 7 global variables and none of them were const. Now it has 4 global variables and 2 of them are const.

in one of my programs. I'm thinking about changing it to:

It's a merger of services and code generation: an on-line code generator that outputs low-level messaging and serialization code based on high-level input. It's implemented as a 3-tier system and uses output from the code generator in each tier. There's also a traditional library that's part of the repo.

Support for more data types for message lengths. Previously message lengths were always 4 bytes. I used this, for example, to reduce the size of the type used for message lengths between the front and middle tiers of the CMW from 4 bytes to 2 bytes.

More concretely, I use it to generate code that's used in each of the tiers mentioned above. The link is to one example of that.