misc-basile

Miscellaneous stuff, mostly single-file tiny programs

Email me to [email protected] for feedback and questions. See starynkevitch.net/Basile for more about me, with this being in English. My pet free software project is RefPerSys.

Some of these files are coded at work, so copyrighted by my employer CEA. My office email is there (before nov. 2023, when I reach retirement) [email protected].

Several single source file programs usually for GNU/Linux/Debian/x86-64. Their compilation command is generally given as a comment inside the source code.

• manydl.c is a program to show that Linux is capable of dlopen-ing many plugins (typically, several hundred thousands or many millions). It works by generating some pseudo-random C file, compiling it into a plugin, which is later dlopen-ed, and repeat.

• forniklas.c is a trivial C program generating then using one single plugin in C. Read its comments for more details.

• redis-scan.c is a program which scans all the keys in a REDIS database (see http://redis.io/ for more) and prints them on stdout.

• execicar.c is a shell-like program interpreting commands on pipes, etc.

• basilemap.ml is a simple exercise to understand the balanced binary trees of the Ocaml stdlib/map.ml file, which I might simplify a bit.

• bwc.c is a crude wc -l like program using getline; for performance benchmarking.

• half.c is a program to stop/cont-inue a command, running it at half load. It was originally written to overcome a hardware bug on an old MSI-270 Turion laptop. It might be useful today to "emulate" bugs by sending periodically an arbitrary signal (such as SIGSEGV or SIGABRT) to the Linux process (or processes) running that command.

• microbenchlist.c is a useless microbenchmark on linked lists use gcc -Wall -O2 -march=native microbenchlist.c -o microbenchlist to compile it.

• makeprimes.c uses the very clever BSD /usr/games/primes program and extract some primes from the stream of primes producing it.

• clever-framac.cc is a clever C++ wrapper for Frama-C static source code analyzer. It uses GNU guile version 3, so the guile-3.0-dev Debian package.

• sync-periodically.c runs periodically the sync(2), is GPLv3+ licensed (so without warranty), and accepts both --version and --help program arguments. It might be useful (to avoid losing a lot of data in kernel file buffers) on power cuts, and don't consume lot of resources. Please glance inside the source code. Your /etc/crontab might have a line like

@reboot sync /usr/local/bin/sync-periodically --log-period=600 --sync-period=3  --daemon --pid-file=/var/run/sync-periodically.pid

and we wrote (for systemd) some sync-periodically.service

• qfontdialog-example.cc is a tiny improvement over this QFontDialog example

• HelloWorld/ contains a small set of files and its README, for a tutorial about GNU make (done on the phone). Perhaps the GPLv3+ license does not fit for such a trivial work.

• onionwebsocket.c is a slighty improved example of websockets from libonion. Most of the code is not mine.

• foldexample.cc is interesting, since it shows how recent C++ compilers are capable of very deep optimizations

• fox-tinyed.cc is for learning the FOX toolkit - a tiny editor (perhaps incomplete).

• fltk-mini-edit.cc is for learning the FLTK toolkit - a tiny editor (perhaps incomplete). It accepts some JSONRPC inspired protocol, documented in file mini-edit-JSONRPC.md.

• logged-gcc.cc is a (GPLv3 licensed) wrapper (coded in C++) around compilation commands by GCC to log them (and their time) with syslog(3) and/or some Sqlite database. You will compile it using compile-logged-gcc.sh. See also this. It could need improvements in start of 2023.

Using logged-gcc

You first need to compile logged-gcc.cc with the compile-logged-gcc.sh shell script. You might want to edit that script. It produces a logged-gcc executable which you could put into your $HOME/bin/ directory.

You then should change your $PATH variable in such a way that $HOME/bin/ is in front of the directory containing your system gcc, usually /usr/bin/. You might have something like export PATH=$HOME/bin:/usr/bin:/bin in some shell initialization file (e.g. $HOME/.bashrc or $HOME/.zshrc for zsh users).

You could run logged-gcc --help to get some help, and logged-gcc --version for version information.

You then add symbolic links with ln -sv $HOME/bin/logged-gcc $HOME/bin/gcc and ln -sv $HOME/bin/logged-gcc $HOME/bin/g++

You could set environment variables $LOGGED_GCC to e.g. /usr/bin/gcc-10 and $LOGGED_GXX to e.g. /usr/bin/g++-10. See also environ(7).

If you want to use logged-gcc with just syslog(3), you don't need to do anything more.

If you want to use logged-gcc with some Sqlite database such as /tmp/logged-gcc.sqlite, you need first to initialize it using logged-gcc --sqlite=/tmp/logged-gcc.sqlite (before any /tmp/logged-gcc.sqlite file exists), and then set the environment variable $LOGGED_SQLITE to /tmp/logged-gcc.sqlite. Only successful GCC compilations go into that database. It is suggested to initialize once then use the default SQLite database $HOME/logged-gcc-db.sqlite ....