And I think it's important to remember that any definition you use is but a choice.
That said, I have never heard of a definition by which dialects differ only by pronunciation: that's the commonest definition of an accent, not a dialect.
I am not, although I have studied languages at university and followed some courses on linguistics and philosophy of language.
I think your definition of language, as in having its own literature, is OK.
@JonasStein What's wrong with the Finance and Home Improvement Stack Exchanges?
I've used HI a bit, and wasn't particularly impressed by it. But I don't have a strong opinion about it. Maybe it would work better for real DIYers. I'm not one of them.
@Fabby, @Cerberus: Another 'Cerberus' for you. He's not what you'd expect, though, and I'm not exactly sure why…
Also, the Conlangery podcast occasionally features/discusses natural languages, if you didn't know about it already. (Not entirely germane to the normal range topics discussed here, but you were talking about it, so meh.)
Wanders off for a bit to do some other stuff before he talks with EliahKagan again, but keeps an eye on the transcript.
@EliahKagan: Let's see if Mailutils v3.4 fixes the build issue I'm trying to capture with my misbehaving logging. Didn't see anything in its upstream release notes that caught my eye as relevant, but it's worth a shot…
Scrolls back up to refresh his memory as to what the other said earlier after the former left last night for a moment…
(Plus some stuff from before I left…)
@EliahKagan Yeah, I just want clean logs so I can report something to Homebrew, but my tools aren't cooperating, the scurvy curs.
@FaheemMitha Home Improvement is on a extreme low level. While I like the topic, I could not read "how can I mount X on a drywall" anymore. Also it is not a typical sx structure. people rather chat there. Finance is similar, but more shocking. ;-)
Gonna be unresponsive for a bit: I'm Fast-User Switching to another account, but will be back to this one as soon as I set up Screen Sharing to it (title case because that's the name of the app on OS X/macOS) through an ssh tunnel.
@RandomDSdevel Hi. My other idea was based on something I think you said but I want to make sure I'm getting this right: when you run your brew command and pipe it to tee, you do get colorized output and at least the appearance of line buffering until./configure scripts start running as part of the build. Is that correct?
./configure scripts, as well as make which is usually run after them, are separate processes from brew that write output. This output may be buffered even if the output of brew is not. This is not really a big idea, just sort of a way of trying to chip away at the problem, to try to figure out how a simpler situation could be created that could reproduce it. However, in connection with something in one of your recent chat messages, it does actually give me an idea...
@RandomDSdevel So I ought to have asked why you were logging the output before (or perhaps you had said why, and I hadn't paid attention). And even aside from that I should have asked about this... but anyway, what I missed then, I can cover now.
When you had talked about piping output, I figured you actually did only want to capture standard output and perhaps just didn't care if stderr were captured (some methods, like script, capture it, but you never mentioned that as an important difference as far as why you weren't using script`). But since you're trying to report a bug, you'll want standard error, too. And I think the absence of standard error in the output might explain some, though not all, of the weirdness you've observed.
It may be that some color codes are being written to the terminal but not your log files. Perhaps even a bug is being masked by color codes being reset in output to standard error. I don't generally expect color codes to be included in such output... but they definitely can be and some programs do include them.
The latter; I believe that the former is only supported in bash 4.x, and OS X/macOS has some version of bash 3.x baked into itself so deeply that I'm not certain how much trouble getting Terminal to use the newer version of the shell, even just for userspace, could cause. And, yes, I've been doing the redirection for tee.
Hmm, okay. Well that idea may be a dead end then. Or not; we may see later. Standard output and standard error usually appear in a well-defined order with respect to one another, because usually a program will flush stdout before writing to stderr, and usually stderr is unbuffered (regardless of output device).
It's possible for a program to behave otherwise, though, and if some child processes run asynchronously and write output, then I would expect that maybe different behavior of the program on the right side of the pipe could affect what text is observed and recorded by changing the order, including the order of escape sequences. But this is all idle guessing now and, so far, I know of now way to idea that idea to attempt solving the problem.
@RandomDSdevel Fortunately, I have another idea. You had mentioned that it would be okay to remove the escape sequences as a separate step, and even that it would be okay to remove them after the log file is written. You can totally do that; it's a completely tractable text processing task, and we have stuff for it here on Unix & Linux:
I can use the "script" command to record an interactive session at the command line. However, this includes all control characters and colour codes. I can remove control characters (like backspace) with "col -b", but I can't find a simple way to remove the colour codes.
Note that I want to use...
Well, I realized that 'the former'/'the latter' can sometimes be ambiguous when referring to items from a list with more than two of them, so I just thought I'd be a tad more specific.
Btw I just checked on an really old Mac that I happen to have in the room with me right now, and you are correct -- Bash 3 does not seem to support |&.
So can you just remove the color codes from the output, either as part of the pipeline (put it after tee and you should still get the colors in what tee writes to your terminal) or afterwards by processing the log file, using the method the answers to that question?
That might not reveal the answer to the puzzle... but it should solve the actual problem.
You had actually linked to that question, but I'm not sure why it wouldn't be sufficient. Also it occurs to me that, at the time you wrote the earlier version of your question, you may not have been considering the possibility of processing the generated logfile afterwards, which I would expect that method either to succeed at or to fail at in a way that could be easily figured out.
@EliahKagan Y'know, I think I know why: the answer to that question that involved scriptreplay and expect was the one that latched on to me, but I couldn't find a version of either the original scriptreplay Perl script or a version of its C port for OS X/macOS, so I dismissed it at the time. I was leaning towards using something that had gone through more cycles of development and testing than one of the Perl scripts on offer there, but perhaps I should just let that go and use one.
I was also thinking that processing the log file afterwards might be problematic since I usually prime it with a line containing a copy of the command that generated the output that ended up in those logs and thought that maybe some of the characters in there could mess processing things after the fact up might not quite work, but I'm not typing any control characters in, so that's probably a moot point anyway.
@EliahKagan Btw, I am sorry I said "put it after tee and you should still get the colors in what tee writes to your terminal" since that's obviously wrong! You want the escape sequences removed from the file that is written, and tee is doing the writing, so filtering out the escape sequences later in the pipe will only keep them from being sent to the terminal (which you don't want to do) and will not remove them from the log (which you do want to do).
Yeah, I switched tabs and checked really quickly since I still had its corresponding question page open. There's an updated version as well, but I don't know whether I really need the changes therein or not (I can probably do without the extra line-ending detection, to start with.)
@EliahKagan Except using tee exposes the run-time buffering problem. I've been able to fix that before, though, just not without also getting escape sequences in my logs.
Well except for the problem with the buffering--I am saying this mainly to correct my earlier error--you can use tee in the opposite way and it should work for this. That is, pass /dev/stdout (macOS should have this device node, I just checked) as the filename to tee, then pipe from tee to the script that removes the escape sequences and redirect that script's output to the log file.
I would actually still suggest trying that, but I would expect that you'd have the same buffering problems.
In any case, Gilles's Perl script or dewtell's extended version of it ought to work just fine on the typescript generated by the script command. Remember that script logs virtually everything you type which typically includes passwords. (Is sudo needed to install things with brew on macOS? I don't remember.)
@EliahKagan OK, I think I'm doing something wrong: what I execute is HOMEBREW_BUILD_FROM_SOURCE=1 script -q /dev/null brew upgrade -vd --build-from-source mailutils 2>&1 | tee -a /dev/stdout | $LOCAL_COPY_OF_ANSI_ESCAPE_SEQUENCE_REMOVING_PERL_SCRIPT >> $LOGFILE. I'm not sure if that's exactly what you meant, though, and may try other combinations (I recall script and tee together working even though bare tee doesn't, so I reverted to that for the moment since it's simpler than…
…the script I was using before.
(Trying to find a link to an online copy of that I found in some SE answer somewhere…)
D'oh, it's in an answer to my question: here it is.
I don't especially recommend using script and tee together. If you're using script, just process the typescript file it generates. You don't need to open /dev/stdout for append with -a, though there should be no harm in doing so. Assuming no word splitting or globbing are happening with $LOCAL_COPY_OF_ANSI_ESCAPE_SEQUENCE_REMOVING_PERL_SCRIPT or $LOGFILE, this should work. I recommend quoting them though.
@Fabby Yeah, his chibi form that he normally stays in is rather cute. If you scroll down, you'll see that the wiki article contains some pictures of his true form as well, and he's much fiercer then.
Eliah: I'm having a really hard time installing Windows and simple Ubuntu on UEFI for the first time, so Arch is going to be for the next time I wipe my system...
@Fabby I'm about half of the way through the entire backlog of episodes, roughly speaking, and the 'cast's rather good so far. You'll probably want to pick and choose which ones you listen to, though.
@Cerberus Honestly, I have no idea. He's not even a dog, either; he's some sort of part-lion chimera thing. No idea why that series's author decided to borrow that name for something that isn't like the original, either.
@Cerberus took me 1 day of finding out how to turn off Secure Boot: you need to password protect the EFI Firmware first before it allows you to turn off secure Boot!
Not that there have been any firmware updates lately. This old jalopy has been around for long enough that its firmware is stable enough. (And I can't run anything newer than OS X v10.11.x on it, anyway, which also contributes to the lack of firmware updates, given the dropped support and all.)
S'cool; you guys probably just missed me while you were talking shop…or at least more technical shop than I usually deal with myself. Not that that isn't 'shop,' either, mind you.
I can use the "script" command to record an interactive session at the command line. However, this includes all control characters and colour codes. I can remove control characters (like backspace) with "col -b", but I can't find a simple way to remove the colour codes.
Note that I want to use...
