mitchweaver / bonsai Goto Github PK

in the readme you say "No .dll hell" when in reality it should be "No dependency hell"

The wiki page you link to, specifically, says "DLL Hell is the Windows ecosystem-specific form of the general concept dependency hell. "

https://github.com/bonsai-linux/bonsai/blob/master/ports/shared/README.md

Output from any X11 program, (xset, xrdb, etc): http://bonsai-linux.org/pastes/uHDIX

With uni starting up again my time is becoming limited.

Any more X11 port work is halted until this problem can be resolved.
If you have any ideas please let me know, thanks.

currently we are using file to determine the compression type and whether is a tar archive, of downloads

This shouldn't be necessary as its given to us via the URL

the CFLAGS=-s and LDFLAGS=-s only work for stripping C-based executables as pointed out by people from the KISS project

To get around this, implement a function to use strip from bin/elf utils

------ also, find out how to get strip, as its not included by default in the toolchains gotten from http://musl.cc

currently ports are just in one big folder, this is already messy at ~60 ports -- can't imagine how cluttered it would be at 1000+

Some distros don't seem to think this is a problem, but I think it is.

Folder support such as core, opt, contrib as the likes of CRUX Linux or a similar style would be nice. Not sure how I want this implemented, just wanted to make this issue to note that is on the radar.

Under Core Philosophy, you state in point 3 about a better c library, "GNU GLibc" which is redundant. glibc is already GNU libc. It's like saying PIN number. You're basically saying "GNU GNU lib c.

This is low priority due to any use of git requiring git as a dependency which is unideal. You should always try to use tarballs instead.

... With that said, I know some people will still want git for "bleeding edge", so it would be nice to add.

I think it would be appropriate to add checksum (probably sha256) as a mandatory variable in pkgfiles. It should be checked against the downloaded file.
Currently, if package's server is hacked, the source file can be modified to be malicious. With checksum, this file is downloaded and then rejected.

should we create a bonsai github organization to store bonsai and bonsai-website? possibly others?

I see that bonsai by itself is taken but bonsai-linux is available, I think that we should use bonsai-linux regardless cause that's what the website is.

example: dropbear

sbase's tar complains that the format is not ustar and errors without extraction

1. Is this fixable? Can we convert to ustar?
2. Is this really the problem at all? Or is it a problem with bunzip2?
3. Is it a bug upstream in sbase?
4. Can we just find a different mirror? Perhaps with working bz2 or gz/xz?

-- I've tried #4, but every mirror is just a literal mirror of the same tarballs with the same problem

• xorg-server
• fontconfig and font stuff
• • libdrm
• • • libpciaccess
• • smdev
• • libepoxy
• • lixau
• • libxdmcp
• • • libbsd
• • libxfont2
• • • freetype
• • • • libbz2
• • • • libpng
• • • libfontenc
• • libxshmfence
• all mesa + gl stuff
• • pixman
• • xkeyboard-config
• • xkbcomp
• • • libx11
• • • • libxcb
• • • libxkbfile
• • xinit
• • • util-linux or just mcookie
• • • xauth
• • • • libxext
• • • • libxmu
• • • • • libxt
• • • • • • libsm
• • • • • • • libuuid
• • • • • • libice
• • • • xmodmap
• • • • xrdb
• • • • • mcpp ??

Use this issue to discuss ports that you wish to be added.

• ghc
• go
• python3
• python2
• openvpn
• dropbear
• neovim
• netcat
• socat
• nfs-utils
• fff <-- (once we get the libedit patch upstream)
• ranger (needs mouse support patched out for netbsd-curses)
• plan9port
• gnupg
• offlineimap
• neomutt
• htop

Always looking for good suggestions for more global flags to add to our .rc

Keep in mind these will be GLOBAL. So they must be sane, (ie not cause breakage).

Post them here for discussion / review.

dependency handling works for adding packages, recursively gathering and adding the needed packages

it would be nice if this were the same for the delpkg() function

Note:

1. only remove packages if they are orphans ie they have NO other packages depending on them
2. Do not remove packages if they are part of a meta package such as core-system, (we dont want to remove our shell, coreutils, etc...)

distro is a placeholder.

Names are hard.

Give me your suggestions

lots of things gracefully fail (on purpose) in the script, however this breaks -e
it would be nice/more secure if this were a possibility.

How to fix:

• wrap statements in if / case
• pipe the 1 exit code to a nop such as:

rmdir _non_existant_ 2> /dev/null ||: