rework libeinfo #651
rework libeinfo #651
Conversation
|
Squashed a few of the commits/cleaned up the messages. |
|
We don't need command line equivalents for ewarnx and eerrorx. |
Uses (n)curses/terminfo for "fancy" stuff. No env vars yet. These functions are not sent to syslog (yet): - ewarn - ewarnx - eerror - eerrorx The syslog id for elog(3) is currently hardcoded. Some intentional API/behavior differences: - ebracket(3) does the same negative column stuff as described in _move_col(). Pretty sick if you ask me. - the length of the "prefix|" string is counted as part of the total length returned by einfo(3) et al.
Properly implement is_quiet(), is_really_quiet(), and is_verbose() by checking the truthyness of the EINFO_QUIET, EERROR_QUIET, EINFO_VERBOSE env vars, respectively. Use the EINFO_LOG env var to set the log id for syslog messages and send ewarn(3), ewarnx(3), eerror(3), and eerrorx(3) to syslog. Only happens if the env var is set tho. Behavior change for EINFO_INDENT: - integer overflow is treated as valid, but >INDENT_MAX (which it arguably is), and therefore chomped to INDENT_MAX API/behavior changes for EINFO_COLOR: - case insensitive on color names - more predictable error case (keep current on parse error) - use the last (valid) definition if multiple given for a single color
Used to ensure that the terminal is actually good for curses use, and enables short circuiting functions that do curses stuff in case the terminal is not.
Replaced with an array and two calls to tput(s). No need for "proper string concatenation" if there is no string concatenation :P
curses/terminfo errors are "basically handled" through: - tiparm() returning NULL on error - tputs() returning an error if the string is NULL We don't actually do anything with the tputs() error since all functions that call it are void.
The rewrite should be "complete" as far as matching features goes. Still gotta do some small TODO's. As well as non-curses.
Set the message in the brackets to "..." if the message is empty. Only possible with ebracket().
Ensure fmt arg is checked at the call site.
|
Please rebase on master. I also see a couple of changes. On line 50 of einfo.h: What does this apply to? Also, can you fix the message to be something like "this is no longer used"? Please update if statements that just have one statement in the if or else blocks to not use curly braces around the statement: should be: Can you also remove the vim mode lines? Thanks, William |
|
I can do the single-line |
| @@ -47,6 +47,7 @@ typedef enum | |||
| } ECOLOR; | |||
|
|
|||
| /*! @brief Returns the ASCII code for the color */ | |||
| __attribute__((__deprecated__("this shit is gone yo"))) | |||
There was a problem hiding this comment.
sorry, but we can't accept things like this. ignoring the puerile nature, it's not helpful or clear to anyone who hits/sees this warning.
There was a problem hiding this comment.
It was merely meant as a message while I was working and not for the final cleaned up/squashed commits.
ok, i flipped through the patch series now, and it doesn't make sense to pull things out. this is a PR showing progress of implementing a new libeinfo library and would make sense squashing all into one. i'm inclined to go with my half-baked idea that i mentioned in the open issue. since we already support building openrc w/out termcap (curses) support, we should lean into that: delete support for termcap entirely and just require an ANSI compliant terminal. |
|
I can close this PR if you instead want to just drop curses and require an ANSI terminal. |
|
you want to send a PR to delete curses support ? |
|
Sure, I think I can do that. |
Rework how libeinfo does curses stuff. This is not by any means ready to merge as-is. For example, OpenRC core has not been changed where needed (places where
ecoloris called are the most obvious ones). There also isn't a non-curses fallback yet, but that should be easy to add once everything else is satisfactory. Not to mention squashing/rewording commits.This is just to get some feedback on what I currently have.
See also: #619