fish's signal handlers are now sufficiently innocuous that there should
be no reason to block signals (outside of temporarily, when creating a
thread and we need to manipulate the signal mask).
Prior to this fix, an "event" was used as both a predicate on which events
to match, and also as the event itself. Re-express these concepts
distinctly: an event is something that happened, an event_handler is the
predicate and name of the function to execute.
Prior to this fix, fish had a signal_list_t that accumulated signals.
Signals were added to an array of integers, with an overflow flag.
The event machinery would attempt to atomically "swap in" the other list.
After this fix, there is a single list of pending signal events, as an array
of atomic booleans. The signal handler sets the boolean corresponding to its
signal.
In a galaxy far, far away, event_blockage_t was intended to block only cetain
events. But it always just blocked everything. Eliminate the event block
mask.
As it turns out, NetBSD's rand(3) is awful - it's possible that in any
given run it'll only return odd numbers, which means
while (rand() % 10)
will never stop.
Since random(3) is also standardized and works, let's use that!
Now that we use an internal process to perform builtin output, simplify the
logic around how it is performed. In particular we no longer have to be
careful about async-safe functions since we do not fork.
Also fix a bunch of comments that no longer apply.
This uses the new internal process mechanism to write output for builtins.
After this the only reason fish ever forks is to execute external processes.
This introduces "internal processes" which are backed by a pthread instead
of a normal process. Internal processes are reaped using the topic
machinery, plugging in neatly alongside the sigchld topic; this means that
process_mark_finished_children() can wait for internal and external
processes simultaneously.
Initially internal processes replace the forked process that fish uses to
write out the output of blocks and functions.
This adds an "in-process" interpretation of dup2s, allowing for fish to
output directly to the correct file descriptor without having to perform
an in-kernel dup2 sequence.
The sigchld generation expresses the idea that, if we receive a sigchld
signal, the generation will be different than when we last recorded it. A
process cannot exit before it has launched, so check the generation count
before process launch. This is an optimization that reduces failing
waitpid calls.
This is a big change to how process reaping works, reimplenting it using
topics. The idea is to simplify the logic in
process_mark_finished_children around blocking, and also prepare for
"internal processes" which do not correspond to real processes.
Before this change, fish would use waitpid() to wait for a process group,
OR would individually poll processes if the process group leader was
unreapable.
After this change, fish no longer ever calls blocking waitpid(). Instead
fish uses the topic mechanism. For each reapable process, fish checks if
it has received a SIGCHLD since last poll; if not it waits until the next
SIGCHLD, and then polls them all.
topic_monitor allows for querying changes posted to one or more topics,
initially sigchld. This will eventually replace the waitpid logic in
process_mark_finished_children().
Comment from the new header:
Topic monitoring support. Topics are conceptually "a thing that can
happen." For example, delivery of a SIGINT, a child process exits, etc. It
is possible to post to a topic, which means that that thing happened.
Associated with each topic is a current generation, which is a 64 bit
value. When you query a topic, you get back a generation. If on the next
query the generation has increased, then it indicates someone posted to
the topic.
For example, if you are monitoring a child process, you can query the
sigchld topic. If it has increased since your last query, it is possible
that your child process has exited.
Topic postings may be coalesced. That is there may be two posts to a given
topic, yet the generation only increases by 1. The only guarantee is that
after a topic post, the current generation value is larger than any value
previously queried.
Tying this all together is the topic_monitor_t. This provides the current
topic generations, and also provides the ability to perform a blocking
wait for any topic to change in a particular topic set. This is the real
power of topics: you can wait for a sigchld signal OR a thread exit.
This resolves the issue where running pre-compiled Linux packages from
binary package manager repositories lead fish to think that we are not
running under WSL.
- Closes#5619.
- Ping neovim/neovim#7330
This happens on OpenIndiana/Solaris/Illumos/SunOS.
Elsewhere we use read_blocked, which already returned in this
case (and which we might want to use here as well!).
`fish_title` as invoked by fish itself is not running in an interactive
context, and attempts to read from the input fd (e.g. via `read`) cause
fish to segfault, go into an infinite loop, or hang at the read prompt
depending on the exact command line and fish version.
This patch addresses that by explicitly closing the input fd when
invoking `fish_title`.
Reported by @floam in #5629. May close that issue, but situation is
unclear.
Taking advantage of the maybe_t's, the logic and nesting here
can be a bit less intense.
Small adjustments to debug output, and found a more accurate
version number for Lion Terminal.app.
Longer term we should have a terminal_t class or something
encapsulating all the kinds of terminal detection we have
with methods that return the color support, and also stuff
like whether the terminal has the newline glitch, the
ambiguous width character behavior, etc.
fish forks child processes when (for example) writing out builtin output.
After fork it resets signal handlers, but if a signal is delivered before
the signal handlers are reset, it will inherit fish's default handlers,
which do things like swallow SIGINT. Teach fish's default signal handlers
to detect this case and re-raise signals with default handlers.
This improves the reliability of control-C in the face of builtins.
I hope this is now complete.
Also, shorten enough descriptions to make `string match --<TAB>`
show a two column pager with 80 cols.
We really should have shown more retraint in the design of `string`,
not all of the flags required both a long and short option created.
300ms was waaay too long, and even 100ms wasn't necessary.
Emacs' evil mode uses 10ms (0.01s), so let's stay a tad higher in case
some terminals are slow.
If anyone really wants to be able to type alt+h with escape, let them
raise the timeout.
Fixes#3904.
`/tmp` isn't present / writeable on every system. Instead of always
using `/tmp`, try to use standard environment variables and
configuration to find a temporary directory.
Adapted from #3974, with updates based on those comments.
Closes#3845.
This is a large change to how io_buffers are filled. The essential problem
comes about with code like (example):
echo ( /bin/pwd )
The output of /bin/pwd must go to fish, not the tty. To arrange for this,
fish does the following:
1. Invoke pipe() to create a pipe.
2. Add an io_bufferfill_t redirection that owns the write end of the pipe.
3. After fork (or equiv), call dup2() to replace pwd's stdout with this pipe.
Now when /bin/pwd writes, it will send output to the read end of the pipe.
But who reads it?
Prior to this fix, fish would do the following in a loop:
1. select() on the pipe with a 10 msec timeout
2. waitpid(WNOHANG) on the pwd proc
This polling is ugly and confusing and is what is replaced here.
With this new change, fish now reads from the pipe via a background thread:
1. Spawn a background pthread, which select()s on the pipe's read end with
a long (100 msec) timeout.
2. In the foreground, waitpid() (allowing hanging) on the pwd proc.
The big win here is a major simplification of job_t::continue_job() since
it no longer has to worry about filling buffers. This will make things
easier for concurrent execution.
It may not be obvious why the background thread still needs a poll (100 msec).
The answer is for cases where the write end of the fd escapes, in particular
background processes invoked inside command substitutions. psub is perhaps
the only important case of this (other shells typically just hang here).
This makes some significant architectual improvements to io_pipe_t and
io_buffer_t.
Prior to this fix, io_buffer_t subclassed io_pipe_t. io_buffer_t is now
replaced with a class io_bufferfill_t, which does not subclass pipe.
io_pipe_t no longer remembers both fds. Instead it has an autoclose_fd_t,
so that the file descriptor ownership is clear.
This switches IO redirections after fork() to use the dup2_list_t,
instead of io_chain_t. This results in simpler code with much simpler
error handling.
This represents a "resolved" io_chain_t, where all of the different io_data_t
types have been reduced to a sequence of dup2() and close(). This will
eliminate a lot of the logic duplication around posix_spawn vs fork, and pave
the way for in-process redirections.
This is a large change to how io_buffers are filled. The essential problem
comes about with code like (example):
echo ( /bin/pwd )
The output of /bin/pwd must go to fish, not the tty. To arrange for this,
fish does the following:
1. Invoke pipe() to create a pipe.
2. Add an io_bufferfill_t redirection that owns the write end of the pipe.
3. After fork (or equiv), call dup2() to replace pwd's stdout with this pipe.
Now when /bin/pwd writes, it will send output to the read end of the pipe.
But who reads it?
Prior to this fix, fish would do the following in a loop:
1. select() on the pipe with a 10 msec timeout
2. waitpid(WNOHANG) on the pwd proc
This polling is ugly and confusing and is what is replaced here.
With this new change, fish now reads from the pipe via a background thread:
1. Spawn a background pthread, which select()s on the pipe's read end with
a long (100 msec) timeout.
2. In the foreground, waitpid() (allowing hanging) on the pwd proc.
The big win here is a major simplification of job_t::continue_job() since
it no longer has to worry about filling buffers. This will make things
easier for concurrent execution.
It may not be obvious why the background thread still needs a poll (100 msec).
The answer is for cases where the write end of the fd escapes, in particular
background processes invoked inside command substitutions. psub is perhaps
the only important case of this (other shells typically just hang here).
This makes some significant architectual improvements to io_pipe_t and
io_buffer_t.
Prior to this fix, io_buffer_t subclassed io_pipe_t. io_buffer_t is now
replaced with a class io_bufferfill_t, which does not subclass pipe.
io_pipe_t no longer remembers both fds. Instead it has an autoclose_fd_t,
so that the file descriptor ownership is clear.
This switches IO redirections after fork() to use the dup2_list_t,
instead of io_chain_t. This results in simpler code with much simpler
error handling.
This represents a "resolved" io_chain_t, where all of the different io_data_t
types have been reduced to a sequence of dup2() and close(). This will
eliminate a lot of the logic duplication around posix_spawn vs fork, and pave
the way for in-process redirections.
By exclusively waiting by pgrp, we can fail to reap processes that
change their own pgrp then either crash or close their fds. If we wind
up in a situation where `waitpid(2)` returns 0 or ECHLD even though we
did not specify `WNOHANG` but we still have unreaped child processes,
wait on them by pid.
Closes#5596.
If we read an R_EOF, we'd try to match mappings to it.
In emacs mode, that's not an issue because the generic binding was
always available, but in vi-normal mode there is no generic binding,
so we'd endlessly loop, waiting for another character.
Fixes#5528.
Originally I sought out to configure the foreground color of the
selected text in the pager. After reading a thread on a github issue I
was inpired to do more: now you can conifgure any part of the pager when
selected, and when a row is secondary. More specifically this commit adds the
ability to specify a pager row's:
- Prefix
- Completion text
- Description
- Background
when said row is selected or secondary.
This will print out along with the stuff we've guessed about color
support. We get a lot of bug reports about these messing up rendering,
this is useful diagnostic output.
Ask the system where utilities are available with confstr (POSIX).
This is the same string printed by `getconf PATH`, which likely
includes more directories.
I was surprised to see:
> set_color normal | string escape
\e\[30m\e\(B\e\[m
I only expected to see a sgr0 here.
Cleanup a nearby `else { if (...) {` and comment with a bogus example.
There was a bogus check for is_interactive_session. But if we are in
reader_readline we are necessarily interactive (even if we are not in
an interactive session, i.e. a fish script invoked some interactive
functionality).
Remove this check.
Fixes#5519
A while loop now evaluates to the last executed command in the body, or
zero if the loop body is empty. This matches POSIX semantics.
Add a bunch of tricky tests.
See #4982
This is effectively a pick of 2ebdcf82ee
and the subsequent fixup. However we also avoid setting WNOHANG unless
waitpid() indicates a process was reaped.
Fixes#5438
For some reason, we have two places where a variable can be read-only:
- By key in env.cpp:is_read_only(), which is checked via set*
- By flag on the actual env_var_t, which is checked e.g. in
parse_execution
The latter didn't happen for non-electric variables like hostname,
because they used the default constructor, because they were
constructed via operator[] (or some such C++-iness).
This caused for-loops to crash on an assert if they used a
non-electric read-only var like $hostname or $SHLVL.
Instead, we explicitly set the flag.
We might want to remove one of the two read-only checks, or something?
Fixes#5548.
Our is_hex_digit() was redundant, we can just use iswxdigit; the libc
implementation is a more efficient table lookup anyhow.
Do is_octal_digit() in terms of iswdigit instead of using wcschr.
This requires threading environment_t through many places, such as completions
and history. We introduce null_environment_t for when the environment isn't
important.
`xlocale.h` is not available on Linux, so we can't just universally
include it.
`HAVE_XLOCALE_H` was already being tested/set in the CMake script as a
possible requirement for `wcstod_l` support, this just adds it to
`config_cmake_h.in` and uses it in `wutil.h` to gate the include.
Removes the dependency on the current user's home directory, instead
overriding it to be within the current hierarchy.
Fixes the tests on Debian buildd, where the home directory is
deliberately unwriteable to pick up errors in builds.
Using `setlocale` is both not thread-safe and not correct, as
a) The global locale is usually stored in static storage, so
simultaneous calls to `setlocale` can result in corruption, and
b) `setlocale` changes the locale for the entire application, not
just the calling thread. This means that even if we wrapped the
`wcstod_l` in a mutex to prevent the previous point, the results
would still be incorrect because this would incorrectly influence the
results of locale-aware functions executed in other threads while
this thread is executing.
The previous comment mentioned that `uselocale` hadn't worked. I'm not
sure what the failing implementation looked like, but `uselocale` can be
tricky. The committed implementation passes the tests for me under Linux
and FreeBSD.
This was the actual issue leading to memory corruption under FreeBSD in
issue #5453, worked around by correcting the detection of `wcstod_l` so
that our version of the function is not called at all.
If we are 100% certain that `wcstod_l` does not exist, then then the
existing code is fine. But given that our checks have failed seperately
on two different platforms already (FreeBSD and Cygwin/newlib), it's a
good precaution to take.
This helps on netbsd, because enter_standout_mode et al are const
there.
These methods don't alter their argument, so they should have been
const to begin with.
This is non-const on macOS, but some of the args we pass are always
const on netbsd.
I have no idea why you'd ever want this to modify its argument, but whatever.
This is the more correct fix for #5447, as regardless of which process
in the job (be it the first or the last) finished first, once we have
waited on a process without ~WNOHANG we don't do that for any subsequent
processes in the job.
It is also a waste to call into the kernel to wait for a process we
already know is completed!
@ridiculousfish had introduced this in 3a45cad12e
to work around an issue with Coverity Scan where it couldn't tell the
mutex was correctly locked, but even with the `fish_mutex_t` hack, it
still emits the same warnings, so there's no pointing in keeping it.
This is necessary for the history race condition test to succeed.
(That test is permanently disabled under WSL (as it always fails) so I
didn't catch this on my end.)
Use `pthread_atfork()` to mark child processes as dirty when `fork()` is
invoked rather than needing to call into the kernel each time
`ASSERT_IS_NOT_FORKED_CHILD()` is called.
This makes simple test cases that hit `ASSERT_IS_NOT_FORKED_CHILD()` 1.8x faster.
------------------------
With a7998c4829 reverted but before this optimization:
```
mqudsi@ZBOOK ~/r/fish-shell> hyperfine -S build/fish 'for i in (seq 100000); test 1 = 1; end'
Benchmark #1: for i in (seq 100000); test 1 = 1; end
Time (mean ± σ): 717.8 ms ± 14.9 ms [User: 503.4 ms, System: 216.2 ms]
Range (min … max): 692.3 ms … 740.2 ms
```
With a7998c4829 reverted and with this optimization:
```
mqudsi@ZBOOK ~/r/fish-shell> hyperfine -S build/fish 'for i in (seq 100000); test 1 = 1; end'
Benchmark #1: for i in (seq 100000); test 1 = 1; end
Time (mean ± σ): 397.2 ms ± 22.3 ms [User: 322.1 ms, System: 79.3 ms]
Range (min … max): 376.0 ms … 444.0 ms
```
Without a7998c4829 reverted and with this optimization:
mqudsi@ZBOOK ~/r/fish-shell> hyperfine -S build/fish 'for i in (seq 100000); test 1 = 1; end'
Benchmark #1: for i in (seq 100000); test 1 = 1; end
Time (mean ± σ): 423.4 ms ± 51.6 ms [User: 363.2 ms, System: 61.3 ms]
Range (min … max): 378.4 ms … 541.1 ms
```
By using a user-land thread-local integer and lock-free (at least under
x86/x64) atomics, we can implement a safe `assert_is_main_thread()`
without calling into the kernel. Thread-local variables are part of
C++11.
This is called a lot in some performance-sensitive areas, so it is worth
optimizing.
This fixes#5438 by having fish block while waiting on a foreground job
via its individual processes by enumerating the procs in reverse order,
such that we hang waiting for the last job in the IO chain to terminate,
rather than the first.
If it's a foreground job, it is related to the currently running exec.
This fixes exec in functions, i.e.
function reload
exec fish
end
would previously always ask about the "function reload" job.
Fixes#5449.
Fixesoh-my-fish/oh-my-fish#664.
Mainly this removes the "TYPE_MASK" macro that just masks off the
higher bits, which I don't think were ever actually used.
Much of this seems like anticipation of future direction, but we're
going somewhere else.
This removes the need to run c-compilation on one file, and allows us
to in future c++-ify this a bit.
There's a lot of bit-fiddling here that is quite unnecessary, better
error-handling would be nice...
So far this removes a few more unused things (because I would have had
to port them), including:
- Functions with ARITY > 3 (even 3 isn't used, but just so we don't
get complacent)
- Variables
- Most functions moved out of the header, because only te_interp is used.
- The te_print function
The function `add_disowned_pgid` adds process *group* ids and not
process ids. It multiplies the value by negative 1 to indicate a wait
on a process group, so the original value must be positive.
If a job is disowned that, for some reason, has a pgid that is special
to waitpid, like 0 (process with pgid of the calling process), -1 (any
process), or our actual pgid, that would lead to us waiting for too
many processes when we later try to reap the disowned processes (to
stop zombies from appearing).
And that means we'd snag away the processes we actually do want to
wait for, which would end with us in a waiting loop.
This is tough to reproduce, the easiest I've found was
fish -ic 'sleep 5 &; disown; set -g __fish_git_prompt_showupstream auto; __fish_git_prompt'
in a git repo.
What we do is to not allow special pgids in the disowned_pids list.
That means we might leave a zombie around (though we probably wait on
0 somewhere), but that's preferable to infinitely looping.
See #5426.
Return STATUS_INVALID_ARGS when failing due to evaluation errors,
so we can tell the difference between an error and falseness.
Add a test for the ERANGE error
The rest of the high-numbered exit codes are not values used by scripts
or builtins, they are internal to fish and come out of
the parser for example.
Prior to adding STATUS_INVALID_ARGS, builtins were usually exiting 2
if they had a special exit status for the situation of bad arguments.
Set it to 2.
We were not parsing an in-range number when we claimed we were,
and were thus failing to error with invalid numbers and returned
a wrong test result. Fixed#5414
Also, provide the detail we can for the other error cases.
Return STATUS_INVALID_ARGS when failing due to evaluation errors,
so we can tell the difference between an error and falseness.
Add a test for the ERANGE error
The rest of the high-numbered exit codes are not values used by scripts
or builtins, they are internal to fish and come out of
the parser for example.
Prior to adding STATUS_INVALID_ARGS, builtins were usually exiting 2
if they had a special exit status for the situation of bad arguments.
Set it to 2.
Cleaned up the code to no longer replicate in fishscript what fish
already does (and caches to boot) in C++ in setting up the paths to the
user configuration directory.
Also introduced a `$__fish_user_data_dir` instead of the sporadic
definitions of `$userdatadir` that may or may not go through
`XDG_DATA_HOME`.
We were not parsing an in-range number when we claimed we were,
and were thus failing to error with invalid numbers and returned
a wrong test result. Fixed#5414
Also, provide the detail we can for the other error cases.
Just sets locale to "C" (because that's the only one we need), does
wcstod and resets the locale.
No idea why uselocale(loc) failed for me, but it did.
Fixes#5407.
This happens in firejail, and it means that we can't use it as an
argument to most pgid-taking functions.
E.g. `wait(0)` means to wait for the _current_ process group,
`tcsetpgrp(0)` doesn't work etc.
So we just stop doing this stuff and hope it works.
Fixes#5295.
This reverts commit 1cb8b2a87b.
argv[0] has the full path in it for a user when he executes it
out of $PATH. This is really annoying in the title which uses $_.
Also check if that is actually defined, not the cur_term proxy.
In #5371, we figured out that there are terminfo entries without this
capability, so this would do a NULL-dereference.
OCLINT was ignoring this, but we can just not do the bad thing.
Declare argc and argv const. These are in the stack, they can
be modified, but we won't.
Fix a typo
... rather than hard code it to "fish". This affects
what is found in $_ and improves the errors:
For example, if fish was ran with ./fish, instead of
something like:
fish: Expected 3 surprises, only got 2 surprises
we'll see:
./fish: Expected 3 surprises, only got 2 surprises
like most other shell utilities. It's just a tiny bit
of detail that can avoid confusion.
This broke fishtape, which did
somestuff | fish -c "source"
Because `source` didn't have a redirection, it refused to read from
stdin.
So, to keep the common issue of `source (command that does not print)`
from seeminly stopping fish, we instead actually check if stdin is a terminal.
This was causing problems if "fish" wasn't in exec_path, like
if the binary had been renamed.
I also noticed that even with 'fish' not renamed, only paths.data
was made relative to my source tree. paths.sysconf, paths.doc, and
paths.bin were all relative to /usr/local.
This had a bunch of "do_{backward,forward}" movements that differed
only in one argument.
Just keep them together, so it's less code, and less needs to be
changed.
If the user is in a directory which has been unlinked, it is possible
for the path .. to not exist, relative to the working directory.
Always pass in the working directory (potentially virtual) to
path_get_cdpath; this ensures we check absolute paths and are immune
from issues if the working directory has been unlinked.
Also introduce a new function path_normalize_for_cd which normalizes the
"join point" of a path and a working directory. This allows us to 'cd' out of
a non-existent directory, but not cd into such a directory.
Fixes#5341
realpath() will return NULL and sets errno if it fails.
We asserted that realpath(".") does not fail. We also didn't really
check that it was successful. Made sure we'll get a perror telling
us about what went wrong if something like this happens again.
Updated tests and added test case
Fixes#5351
If fish detects that it was started with a pgrp of 0 (which appears to
oddly be the case when run under firejail), create new process group for
fish and give it control of the terminal.
This selectively reverts 55b3c45 in cases where an invalid pgrp is
detected. Note that this is known to cause problems in other cases, such
as #3805 and Microsoft/WSL#1653, although the former may have been
ameliorated or even addressed by the recent job control overhaul, so
that's why we are careful to only assign fish to its own pgroup if an
invalid pgroup was detected and not as the normal case.
This reverts commit 54050bd4c5.
Type job_list_t was changed from a list to a deque in
commit 54050bd4c5.
In process_clean_after_marking(), we remove jobs while iterating.
dequeues do not support that. Make it a list again.
This fixes the `~floam/` case, where the out_tail_idx pointer needs to
point to the "/", not the last letter.
The `~/` and `~floam` cases still work.
Unfortunately, I'm unsure of how to test this.
Fixes#5325.
Limit the fish_wcstod fast path to ASCII digits only, to fix the problem
observed in the discussion for a700acadfa
where LANG=de_DE.UTF-8 would cause `test` to interpret commas instead of
periods inside floating point values.
Now jobs are aware of their parent jobs, and can interrogate those jobs,
to determine if every job in the chain is fully constructed.
Remove flags and the static stacks that manipulated them.
The parent of a job is the parent pipeline that executed the function or
block corresponding to this job. This will help simplify
process_mark_finished_children().
Prior to this fix, cding into a symlink and then completing .. would complete
from the physical directory instead of the logical directory, which could not
actually be cd'd to. Teach cd completiond to use the logical directory.
select_try() returned IO_ERROR to indicate that there's no file descriptors
from which to read. Name this return value properly.
Also migrate this type into proc.cpp since it's not used outside of the
header.
This is an opposite case from the usual "pipe into grep-the-function"
where my `pbpaste` emitted a lot of content exceeding the OS pipe
buffer. The `block_on_fg` condition was just `send_sigcont` in the
original job control rewrite, and it was incorrect to sub it for
WAIT_BY_PROCESS on its own.
However, this requires always blocking when select_try returns an
interrupted/incomplete read or else fish doesn't block and stays running
in a tight loop in the background (and incorrectly writing to a terminal
it doesn't own under higher debug levels), which I *think* is OK.
Instantiate the std:locale instance used within the character comparison
callback outside the lambda and take a reference to it instead of
creating the locale object for each character in the sequence.
This is part of a very tight loop with lots of inputs during the
evaluation of fuzzy string matches for completions/autosuggestions and
is worth optimizing.
This was introduced in 1b1bc28c0a but did
not cause any problems until the job control refactor, which caused it
to attempt to signal the calling `exec` builtin's own (invalid) pgrp
with SIGHUP.
Also improved debugging for `j->signal()` failures by printing the
signal we tried sending in case of error, rename the function to
`hup_background_jobs`, and move it from `reader.h`/`reader.cpp` to
`proc.h`/`proc.cpp`.
When a function is encountered by exec_job, a new context is created for
its execution from the ground up, with a new job and all, ultimately
resulting in a recursive call to exec_job from the same (main) thread.
Since each time exec_job encounters a new job with external commands
that needs terminal control it creates a new pgrp and gives it control
of the terminal (tcsetpgrp & co), this effectively takes control away
from the previously spawned external commands which may be (and likely
are) expecting to still have terminal access.
This commit attempts to detect when such a situation arises by handling
recursive calls to exec_job (which can only happen if the pipeline
included a function) by borrowing the pgrp from the (necessarily still
active) parent job and spawning new external commands into it.
When a parent job spawns new jobs due to the evaluation of a new
function (which shouldn't be the case in the first place), we end up
with two distinct jobs sharing one pgrp (to fix#3952). This can lead to
early termination of a pgrp if finished parent job children are reaped
before future processes in either the parent or future child jobs can
join it.
While the parent job is under construction, require that waitpid(2)
calls for the child job be done by process id and not job pgrp.
Closes#3952.
Use SIGCHLD to determine whether or not waitpid(2) calls can be elided,
but only with extreme caution. If we receive SIGCHLD but are not able to
reap all jobs, we need to iterate through them again.
For this to work, we need to make sure that we reap all children that we
can reap after a SIGCHLD, i.e. it's not OK to just reap the first and
return or else we can never clear the dirty state flag.
In all cases, as expensive as a call to waitpid() may be, if a child
process is available for reaping it is always cheaper to wait on it then
reap it than to call select_try() and end up timing out.
The old code was rather haphazard with regards to error control, and
would make mutable changes before operations that could fail without any
viable error handling options.
Convert `select_try()` to return a well-defined enum describing its
state, and handle each of the three possible cases with clear reasons
why we are blocking or not blocking in each subsequent call to
`process_mark_finished_children()`.
* Use the newly-introduced signal_block_t RAII wrapper
* Remove EINTR loops as all signals are blocked
* Clean up control flow thanks to RAII wrappers
* Rename parameter to clarify what it does and update docs accordingly
* Update outdated comments referencing SIGSTOP code that was removed a
long time ago.
* Remove no-op CHECK_BLOCK() call
