| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
| |
Fixes: ea4fb2a1e34900635d0db7bd1d8e5e1bea06ced0
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
| |
Obsolete as of bfd5710de9b14712c4a621259b951bd10489d0c0. A neat tool, but
it ended up not being a good fit for the reasons described in that commit.
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
| |
No longer needed since bfd5710de9b14712c4a621259b951bd10489d0c0.
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
| |
Kerin fixed this in a later push but I'd already pulled it in by that
point.
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Following almost a year of usage and testing, I now consider my
experiments in smart terminal handling to be both a success and a
failure by equal measure. I deem them to be a successful in so far as
the code functions as intended, and that there have been several
positive reports from those intrepid enough to unmask
gentoo-functions-1.0. However, I also deem them to be a failure in so
far as there is a noticeable impact upon performance. In particular,
repeated calls to ebegin() and eend() are handled rather slowly. As
such, this commit greatly simplifies the code whilst retaining the most
compelling traits of the work that has been performed. The exact changes
are described herewith.
The _eprint() function shall now only check whether STDOUT refers to a
tty. In the event that it does, the message may be printed in a
colourised fashion. Otherwise, it shall be printed plainly. No longer
will the terminal be asked to remember the position of the cursor (using
DECSC), nor shall any variables be used for the purpose of keeping
state.
The _eend() function shall no longer attempt to restore the prior
position of the cursor (using DECRC), nor attempt to determine the
present position of the cursor (using ecma48-cpr). Instead, the
behaviour is reduced to that of checking whether STDOUT refers to a
smart terminal. In the event that it does, the success/failure
identifier shall always be printed on the preceding line, right-aligned.
Otherwise, it shall be printed plainly on the present line. It remains
the case that the terminal will be asked to report its dimensions, so
that _eend() can adapt to any adjustments made to the width of the
terminal emulator.
As a consequence of these changes, the _update_cursor_coords() and
_ecma48_cpr() functions are rendered obsolete and have been removed.
Additionally, the "genfun_cols", "genfun_y" and "genfun_is_pending_lf"
variables are rendered obsolete and have been removed. Finally, the
_update_winsize() function has been renamed to _update_columns().
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
| |
For it is an edge case and it is not clear that it is worth it. Instead,
check at the time of sourcing. The underlying bug really ought to be
addressed upstream.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
| |
I tend to define functions in accordance with the K&R brace style. In
deference to the traditions of the project, employ the Allman style
instead.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
| |
In Emacs, M-x term opens an "eterm-color" terminal, whose
implementation of the CHA (ECMA-48 CSI) sequence suffers from an
off-by-one error. Work around it. Note that Eterm does not have this
same bug.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In this case, the offender is the PTY that is set up by the
_create_pty_or_pipe() utility function in portage, which uses python's
pty module. I transpires that its behaviour is not exactly equivalent to
whichever controlling terminal is in effect at the time that emerge(1)
is executed, if any. As a consequence, the ecma48-cpr utility times out
while waiting for a response to the CPR (ECMA-48 CSI) sequence.
I can discern no option other than to refrain from considering the
terminal as being "smart" in this case. This is accomplished by
recognising that stty(1) reports "0 0" when asked to check its size.
These are nonsensical readings that would never be produced for a
properly functioning, non-dumb terminal. In fact, this check used to be
in place but was removed, simply because I forgot how important it is.
Processes running under portage that call into gentoo-functions will
still be able to produce coloured messages, but all the advantages of
the smart terminal handling path will be obviated. This is unfortunate,
but I have no intention of returning to the original, broken behaviour
of gentoo-functions, which was to assume that the characteristics of the
terminal found at the time of sourcing hold forever, and that STDOUT
and STDERR will never refer to anything other than that same terminal.
I would add that script(1) does not exhibit the same issue.
$ tty
/dev/pts/0
$ script -c 'tty; stty size <&1; /lib/gentoo/ecma48-cpr <&1' /dev/null
Script started, output log file is '/dev/null'.
/dev/pts/11
34 144
34 1
Script done.
In other words, it is portage that is broken and which should be fixed.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Reported-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
| |
There seems little sense in honouring these signals at the point that
the terminal settings are about to be altered and the CPR sequence
written.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
| |
While at it, don't bother disabling ECHOE, ECHOK and ECHONL, for they do
nothing in the case that ICANON has been disabled.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
| |
The idea being to narrow the window of time during which the terminal
is in its raw mode as much as is possible.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
| |
Doing so saves an additional syscall and waits for the output queue to
be drained, which is no bad thing.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
| |
The test determines whether ebegin() appends at least one <newline>
character to the message (following the ASCII ellipsis).
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
|
|
|
| |
The CUP (ECMA-48 CSI) sequence expects for the row to come first, and
the column second. I could have just swapped the arguments but, instead,
I have adjusted the surrounding code so as to strictly adhere to this
convention (one that is also observed by stty size). This should
eliminate the possibility of any such mistake being made in the future.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Fixes: 20bc15b5b1009149c4a3d531911d3c219dc55f3a
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The ebegin() function was recently modified to strip the trailing <newline>,
if any, from the given message. Later, it was modified to strip all trailing
<newline> characters. The purpose of doing so is to ensure that there is
no accidental line breaking betwixt the message and the ASCII ellipsis
that is appended. However, the function should then have been appending a
single <newline> to the composed message before handing it off to _eprint().
Make it so.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Fixes: 1947f0ed81f3b95a7e10a8a5a707776948f8a487
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous "fix" was an experimental patch that was not yet intended
for inclusion. This supplemental commit does as little as is necessary
to render the test suite properly functional under meson, given that
meson.build has been configured to expect the use of the TAP protocol.
It also configures meson to be verbose while executing the test suite.
As far as I can gather, the TAP support in meson is weak. Subtests do
not appear to be supported. For that reason, I have dropped the TAP
version from 14 to 13. While meson recognises "# SKIP" as a directive,
it does not report the reason that may follow the text. Another issue is
that no controlling terminal is present under meson. Therefore,
test_update_cursor_coords() is rendered useless, for it will always be
skipped. I have left the test in for the time being, in the hope that
there may yet be a solution.
In any case, this should be enough to render the -9999 ebuild both
testable and usable.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
| |
Given that there are no subtests, it isn't permissible to duplicate any
of the test numbers. Also, the test_update_cursor_coords() function was
not correctly conveying the case where the test is skipped.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
|
| |
False-positives galore. It's getting out of hand for test-functions, so
I disabled several more tests in its global scope. It was correct to
point out that "${PWD}" should be quoted, however. Bash doesn't mind but
other sh implementations might.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
| |
I changed my mind about this. It's probable that more diagnostic
messages will be added in the future and it may then be useful to
observe them in the case that the test suite fails.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
| |
That this was not happening did not yet amount to a bug but let's ensure
that it cannot be one in the future.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
| |
I am fed up with the indentation behaviour being wrecked by the
imablance between ts and sw and I aleady assume that a tabstop is
equivalent to 8 spaces. Enough.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
| |
Have ebegin() strip all trailing newlines rather than just one at most.
Likewise for _eprint() in the case that a smart terminal is found.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
|
|
|
| |
There is no guarantee that the input queue is empty at the time that
ecma48-cpr is executed. If the terminal has input waiting, not only
will ecma48-cpr potentially fail but it will pollute the input queue.
Work around this by flushing the input queue entirely. I'm not
especially pleased by this approach but I can see no other option,
short of giving up on the idea of using the CPR sequence at all.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
| |
The test phase is run prior to the install phase. Compensate for this by
wrapping ecma48-cpr with a shim function and redeclaring said function
where the value of EBUILD_PHASE is equal to "test".
Signed-off-by: Kerin Millar <kfm@plushkava.net>
|
| |
|
|
|
|
|
|
|
|
| |
Right now, I don't think there's much benefit for putting the ecma tool
in libexecdir given it's in /usr and everything assumes functions.sh is at
/lib/gentoo/functions.sh - i.e. it unnecessarily breaks split-usr.
Maybe we can just add some symlinks later though.
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
... yielding
```
/tmp/destdir
└── usr
└── local
├── bin
│ └── consoletype
├── lib
│ └── gentoo
│ └── functions.sh
├── libexec
│ └── gentoo
│ └── ecma48-cpr
└── share
└── man
└── man1
└── consoletype.1
11 directories, 4 files
```
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
To run cd "$dir" is problematic because:
1) it may consider its operand as an option
2) it will search CDPATH for an operand not beginning with ./, ../ or /
3) it will switch to OLDPWD if the operand is -
4) cdable_vars causes bash to treat the operand as a potential varname
This commit introduces a chdir() function that addresses all of these
pitfalls.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit introduces a function named _update_tty_level(), whose
reponsibility is to grade the capability of the terminal, if any, on a
scale of 0 to 2. A value of 0 indicates that no terminal has been
detected, 1 that a dumb terminal has been detected, while 2 indicates
that a smart terminal has been detected. For the terminal to be
considered as smart, all of the following conditions must hold true.
- the TERM variable must not contain "dumb" as a substring
- the stty(1) utility must be able to report the terminal's dimensions
- the cursor's position must be reported for the ECMA-48 CPR sequence
Both the _eprint() and _eend() functions have been adjusted, so as to
evaluate the terminal's capabilites on each occasion that they are
called, with the overall intention of better exploiting the capabilities
of modern terminals, thus providing a better experience for a majority
of Gentoo users. These adjustments are described herewith.
Common to both functions is that the mechanism of tracking whether a
<newline> character is pending for the last printed message is no longer
employed for smart terminals.
As concerns _eprint(), it now begins by not only determining whether a
terminal is attached to STDOUT, but whether it is a smart terminal. In
the case that it is smart, it shall be determined whether the cursor is
situated at the first column. Should it not be, a <newline> character
shall be emitted, on the basis that the last call may have been to a
function that printed a message without a trailing <newline>.
Next, the message shall be checked, so as to determine whether it
contains a trailing <newline> character. Should it not, the message
shall be printed, along with a trailing DECSC sequence. The effect of
this sequence is to instruct the terminal emulator to save the current
position of the cursor.
If, on the other hand, the message is found to contain a trailing
<newline> character, it shall be stripped prior to printing the message.
The terminal shall then be asked to report the new position of the
cursor. Should the cursor turn out to be situated at the very last row
of the terminal, it shall be assumed that vertical scrolling is about to
occur and the cursor shall be moved up by one row, prior to its position
being saved by way of the DECSC sequence. Thereafter, two <newline>
characters shall be printed, so as to advance to the next line (relative
to the message). Otherwise, should the cursor be found to be situated at
a row that precedes the last row of the terminal, the DECSC sequence
shall be emitted, followed by a single <newline> character.
Note that the overall intention is to stash the exact position of the
cursor at the point that it is situated at the end of the printed
message, but just before a <newline> is printed (if any).
As concerns _eend(), and as with _eprint(), it now begins by not only
determining whether a terminal is attached to STDOUT, but whether it is
a smart terminal, though not before saving the last-known dimensions
of the terminal. Next, in the case of a dumb terminal - or no terminal
at all - the message shall be printed with a single preceding space.
That is, no attempt shall be made to right-align the success/failure
indicator. If, on the other hand, the terminal is found to be smart, the
behaviour shall be as described by the next paragraph.
The current dimensions of the terminal shall be compared against the
last-known dimensions of the terminal, as they were prior to the
terminal grading procedure. The purpose of doing so is to determine
whether the terminal has been recently resized. For the terminal not to
have been resized implies that it may yet be possible to write the
success/failure indicator on the very same row as the last printed
message, even if it were <newline>-terminated. With that in mind, the
current position of the cursor shall be saved. Next, the DECRC sequence
shall be emitted to the terminal. The effect of this sequence is to
instruct the terminal emulator to restore the position of the cursor,
as was saved by the DECSC sequence. In other words, the cursor position
shall be moved to wherever it was just after the last message was
printed, albeit before the trailing <newline> character (if any). Next,
the terminal shall be asked to report the new position of the cursor. In
the event that the cursor is found to have moved to any other row than
the immediately preceding one, it shall be assumed that scrolling has
occurred since printing the last message, in which case the cursor shall
be moved back to its prior position. Otherwise, the cursor shall be left
at its present position.
With that out of the way, the degree of indentation required to
right-align the indicator shall be trivially calculated by deducting
the present column number from the total number of columns spanned by
the terminal, with a further 6 deducted to account for the width of the
indicator. Should the resulting value be greater than 0, the cursor
shall be shifted that many cells to the right. Should the resulting
value be negative, it shall be recognised that there is insufficient
room to print the indicator on the same line and, instead, the indicator
shall be printed on the next line. Should the resulting value be 0, it
shall be recognised that the cursor is already situated at the correct
column.
Finally, the indicator shall be printed, along with a trailing <newline>
character.
As a consequence of these changes, the genfun_endcol variable has been
eliminated, as it is no longer required.
It should be noted that both DECSC and DECRC are private mode sequences,
and are, thus, not defined by ECMA-48. However, they were introduced by
DEC for the VT100 terminal, which was released in 1978. As such, they
can be considered as a de-facto standard.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The ECMA-48 specification defines the CPR (Cursor Position Report)
sequence which, as its name implies, instructs the terminal to divulge
the present coordinates of the cursor. Though it is theoretically
possible to make use of this sequence by using the shell and standard
utilities, it would be tremendously costly and somewhat unreliable to
do so. This commit introduces a utility - written in C - that does the
job.
Its behaviour is quite simple. If STDIN is found to be a terminal, an
attempt will be made to initiate - and read - a Cursor Position Report.
Upon success, the present row and column of the cursor shall be printed
as two space-separated decimal integers. Otherwise, a diagnostic
message shall be printed to STDERR, and the utility shall exit with
a non-zero status.
This utility will be used by the impending overhaul of the _eprint()
and _eend() functions.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
| |
The is_older_than() function now calls ewarn() if given invalid
arguments. For the purposes of the test suite, the resulting diagnostic
messages aren't particularly interesting.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
| |
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This new function assumes responsibility for updating the genfun_cols
variable. Additionally, it assumes responsibility for updating the newly
introduced genfun_rows variable.
As before, stty(1) is used to determine the dimensions of the terminal.
The approach has been slightly altered so as to incur just one subshell
in the course of doing so. The reason for this is that changes will soon
be made to the _eprint() and _eend() functions that require for the
function to be called repeatedly, not merely at the time of sourcing
functions.sh.
Should stty(1) fail - or produce nonsensical output - it will no longer
be assumed that there are 80 available columns. The requirement for the
"[ ok ]" and "[ !! ]" indicators to always be indented was recently
eliminated, so there is no longer any need to falsify a value for
genfun_cols.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
While there is nothing wrong with the way in which it is presently
being used, the feature is sufficiently bug-ridden that it will not work
correctly with the impending overhaul of the _eprint() and _eend()
functions. As such, drop support for it in advance. Chet has pledged to
treat my bug report as a feature request, so it may yet be possible to
re-introduce support for it in the future.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Bug: https://lists.gnu.org/archive/html/bug-bash/2023-02/msg00142.html
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
| |
This is a precautionary measure, given that ebegin() appends " ..." to
the message.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The genfun_lastbegun_strlen variable was previously used by _eend() to
indent the "[ ok ]" and "[ !! ]" indicators in the case that STDOUT is
found not to be a tty. Dispense with this variable. Instead, refrain
from indenting the indicator at all. After all, the width of the
controlling terminal is immaterial unless one is actually intending to
print to it.
$ { ebegin "Testing"; eend 0; } | cat
* Testing ... [ ok ]
Apart from simplifying the code, this change brings the overall
behaviour of the printing functions closer to that of their counterparts
in OpenRC.
The genfun_lastcall variable was previously used by the _eprint()
function for the sole purpose of printing a LF (newline) character in
the case that the last message was printed as a consequence of calling
the ebegin() function. It would do so in anticipation of the _eend()
function later emitting the CUU (ECMA-48 CSI) sequence to move the
cursor up by one line, just prior to printing the "[ ok ]" and "[ !! ]"
indicators. Additionally, it was used by the ebegin() function to
determine whether a terminating LF character should follow the printed
message. Specifically, it would elect to print a LF character in the
case that ECMA-48 CSI sequences had been disabled at the time of
functions.sh being sourced. Finally, the value of genfun_lastcall would
influence the (now defunct) method by which the _eend() function would
calculate the degree of indentation required for the indicator in the
case of STDOUT not being a tty.
This variable has been dispensed with and replaced by a variable named
genfun_is_pending_lf. As its name suggests, its purpose is to track
whether the last printed message happened to contain a terminating LF
character. Now, whenever the _eprint() function is called, it consults
the variable so as to determine whether a LF character should be printed
for the purpose of terminating the last message, just before proceeding
to print the next one. Once the next one has been printed, the value of
the variable is updated accordingly. Similarly, the _eend() function
consults the variable so as to determine whether a LF character should
be printed to a terminal, just prior to the CUU (ECMA-48 CSI) sequence.
Ultimately, ebegin() will no longer be sloppily treated as a special
case. Rather, any printing function that inhibits the addition of a
terminating LF character (ebegin, einfon, ewarn, errorn etc) will have
its message be terminated upon calling any printing function thereafter.
In addition to cleaning up the code a little, these changes should
render the impending overhaul of the _eprint() and _eend() functions
easier to digest, once it lands.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Both the _eprint() and _eend() functions will be overhauled in the near
future, at which point conveying diagnostic messages (indirectly) to
_eprint() will become beneficial. This commit affects the following
functions.
- esyslog
- is_older_than
- veend
- vewend
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Up until very recently, functions.sh would not attempt to determine
whether a given message was being printed to a terminal. Over the years,
this has caused some deleterious effects, such as printing ECMA-48 CSI
and SGR sequences to file descriptors that are not necessarily attached
to a terminal. Another issue used to be that these sequences would be
emitted to terminals identifying themselves as "dumb".
Rather than tackle the underlying problems, the historical response has
tended towards further entrenching the use of the ncurses implementation
of tput(1). While it has its uses, the way in which it is employed is in
no way portable. POSIX only guarantees the availability of the clear,
init and reset operands. Presently, the colors operand is used as a
heuristic to determine whether the ncurses implementation is available
before proceeding to run a series of commands that will usually output
ECMA-48 SGR sequences that are standard for all but ancient and/or
obscure video terminals.
Another present use of tput(1) is in generating the CUU1 and CUF
sequences. In the overwhelming majority of cases, it will generate these
as standard ECMA-48 CSI sequences. To put this in perspective, the DEC
VT100, a video terminal released almost 45 years ago, supports these
sequences. The ECMA-48 specification, itself, is almost as old.
Unfortunately, reasoning with such matters tends towards a dichotomy.
On the one hand, there is a camp that considers it a cardinal sin to
attempt to emit an ANSI escape sequence without first having consulted a
specific implementation of tput(1). On the other, there is the camp that
points out that ECMA-48 sequences are highly portable and that video
terminals from the 1970s and 1980s are largely obsolete. Here are two
articles that present opposing points of view.
https://mywiki.wooledge.org/BashFAQ/037
https://xn--rpa.cc/irl/term.html
Of late, I find the arguments presented by the latter camp to be
increasingly convincing. As such, the goal of this commit is twofold.
Firstly, to implement an independent method for detecting a dumb
terminal. To that end, a _has_dumb_terminal() function has been added,
which simply checks the value of TERM. This, alone, determines whether
the CUU1 and CUF sequences should be assigned, with tput(1) no longer
being used to generate them.
Secondly, to further dial back on the use of tput(1) by not using it to
generate the SGR (colour) sequences. While I believe that nobody would
notice if we were to do away with the use of tput(1) altogether, I have
taken a nuanced approach by continuing to use it for detecting the
number of available colours. To that end, a _has_monochrome_terminal()
function has been implemented. This, alone, determines whether the SGR
sequences should be assigned. The function works by first checking
whether the terminal is dumb. If not, it tries to run "tput colors",
before checking whether -1 colours are reported, for that is how the
ncurses implementation reports the absence of colour support.
The resulting code is simpler and more pleasing to read, with tput being
run once, at most. Also, because there is no longer a requirement to
handle the CSI sequences in a raw form, they are now declared as strings
containing backslash-escape sequences, with printf %b being used to
decode them. This has the advantage of not emitting the raw codes to
STDERR while using xtrace for debugging.
In the future, I think that it may well be possible to drop the function
that detects a monochrome terminal and rely on dumb terminal detection
instead but let's see how it goes.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
| |
|
|
|
|
|
|
|
| |
A beneficial effect of doing so, apart from being able to show one more
character, is that the the associated arithmetic expansions uniformly
work with the genuine length of the end column, which is 7.
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
|
Sam James
Kerin Millar
GitWeb