Extra terminal modes breaks build on FreeBSD

[mboes]

The fix to #98 breaks the build on FreeBSD. In particular, many of the extra constants introduced in the patch have no declaration C side on FreeBSD. This in turn means that GHC HEAD does not build on FreeBSD currently.

A workaround is to define these constants manually in System.Posix.Terminal.Common:

#ifdef __FreeBSD__
#define BSDLY  0020000
#define IUCLC  0001000
#define OFILL  0000100
#define OFDEL  0000200
#define NLDLY  0000400
#define BS0    0000000
#define BS1    0020000
#define CR0    0000000
#define CR1    0001000
#define CR2    0002000
#define CR3    0003000
#define CRDLY  0003000
#define FF0    0000000
#define FF1    0100000
#define FFDLY  0100000
#define NL0    0000000
#define NL1    0000400
#define TAB1   0004000
#define TAB2   0010000
#define VT0    0000000
#define VT1    0040000
#define VTDLY  0040000
#if OSVERSION < 800000
#define TAB0   0000000
#define TAB3   0014000
#define TABDLY 0014000
#endif
#endif

Not great. But I didn't just invent this patch. I stole it verbatim from the comms/remserial port.

Another option is to conditionally not compile these modes on platforms that don't support them. But that would be terrible. It would cause the same type of problems as we already have in the network package.

cc @bgamari @mchakravarty

[hvr]

sigh this was what I was worried about in #99, but I didn't suspect FreeBSD to be the one causing troubles :-/

[geekosaur]

This is not that surprising to me. None of the hard-delay stuff should be applicable to anything but printing terminals and the very oldest original glass ttys. (Some more recent stuff is slow enough to need something delay-like, notably the original VT100, but those all speak flow control as well and it's a better solution in general because it's adaptive.)

[Fuuzetsu]

What are we going to do about this? Can it be merged in?

[hvr]

@Fuuzetsu to be honest, if @geekosaur is right, and nobody should nowadays have need to for those, I'd be tempted to remove them again, rather than fake them.

[glguy]

@bbarenblat Seemed to need them for something and that lead us to #98 in the first place. I'd be interested to learn what ONLCR was needed for in that case.

[hvr]

Yes, and ONLCR doesn't seem to be missing on FreeBSD, so it'd be fine to keep that one...

[geekosaur]

ONLCR is standard. OCRNL would be the weird one.

…

On Thu, Oct 26, 2017 at 3:33 PM, Herbert Valerio Riedel < ***@***.***> wrote: Yes, and ONLCR doesn't seem to be missing on FreeBSD, so it'd be fine to keep that one... — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#102 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AB8SoPZQalgkA5NtLVmvV0e3G4JqHGmcks5swN6lgaJpZM4PxxeE> .

-- brandon s allbery kf8nh sine nomine associates allbery.b@gmail.com ballbery@sinenomine.net unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

[bgamari]

Let's rip out OCRNL, adding a comment explaining the trickyness that it introduces.

[hvr]

@bgamari Now I'm confused; I was under the impression we need to revert/rip-out everything but ONLCR (which is the only thing that was actually demanded in #98, but we went beyond and added all missing POSIX-declared entities while at it) that was added in #99. I.e. it's not enough to just remove OCRNL, but we need to rip everything else mentioned in the CPP in #102 (comment)

[bbarenblat]

Are you sure ONLCR or OCRNL is not defined? FreeBSD’s termios(4) describes ONLCR, OCRNL, ONOCR ONLRET, TABDLY(TAB0), and TABDLY(TAB3). I don’t have a FreeBSD box to test with, but does this work if you rip out OFDEL, OFILL, NLDLY, CRDLY, TABDLY(TAB{1,2}), BSDLY, VTDLY, and FFDLY?

I’m implementing a terminal emulator, so many of the CR/LF mapping options (and TAB3, which expands tabs to spaces) are potentially useful. I am already using OCRNL to move some output normalization out of my code and into the PTY subsystem.

[hvr]

Soon I'll get myself a VM with FreeBSD to be able to investigate this first-hand ;-)

[geekosaur]

If I am understanding this correctly, that one was not the issue; it's the delays, printing terminal modes, and such that were added at the same time that cause compatibility issues. They exist on Linux (but mostly unimplemented/ignored) and on commercial Unixes but not the *BSDs.

…

On Mon, Oct 30, 2017 at 8:56 AM, Benjamin Barenblat < ***@***.***> wrote: Are you sure ONLCR or OCRNL is not defined? FreeBSD’s termios(4) <https://www.freebsd.org/cgi/man.cgi?query=termios&sektion=4> describes ONLCR, OCRNL, ONOCR ONLRET, TABDLY(TAB0), and TABDLY(TAB3). I don’t have a FreeBSD box to test with, but does this work if you rip out OFDEL, OFILL, NLDLY, CRDLY, TABDLY(TAB{1,2}), BSDLY, VTDLY, and FFDLY? I’m implementing a terminal emulator, so many of the CR/LF mapping options (and TAB3, which expands tabs to spaces) are potentially useful. I am already using OCRNL to move some output normalization out of my code and into the PTY subsystem. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#102 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AB8SoJ0RaqLmq3iV3lYaeoQr3bzqz896ks5sxcdjgaJpZM4PxxeE> .

-- brandon s allbery kf8nh sine nomine associates allbery.b@gmail.com ballbery@sinenomine.net unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

[bgamari]

Let's rip out OCRNL, adding a comment explaining the trickyness that it introduces.
@bgamari Now I'm confused; I was under the impression we need to revert/rip-out everything but ONLCR

Yes, ignore my first comment; it clearly wasn't well-considered.

I have tried building unix 748e8b4 on FreeBSD 11 and found the following are missing:

In file included from dist/build/System/Posix/Terminal/Common_hsc_make.c:1:0:
Common.hsc: In function 'main':
Common.hsc:225:16: error: 'OFDEL' undeclared (first use in this function)
Common.hsc:226:16: error: 'OFILL' undeclared (first use in this function)
Common.hsc:227:16: error: 'NL0' undeclared (first use in this function)
Common.hsc:228:16: error: 'NL1' undeclared (first use in this function)
Common.hsc:229:16: error: 'CR0' undeclared (first use in this function)
Common.hsc:230:16: error: 'CR1' undeclared (first use in this function)
Common.hsc:231:16: error: 'CR2' undeclared (first use in this function)
Common.hsc:232:16: error: 'CR3' undeclared (first use in this function)
Common.hsc:234:16: error: 'TAB1' undeclared (first use in this function)
Common.hsc:235:16: error: 'TAB2' undeclared (first use in this function)
Common.hsc:237:16: error: 'BS0' undeclared (first use in this function)
Common.hsc:238:16: error: 'BS1' undeclared (first use in this function)
Common.hsc:239:16: error: 'VT0' undeclared (first use in this function)
Common.hsc:240:16: error: 'VT1' undeclared (first use in this function)
Common.hsc:241:16: error: 'FF0' undeclared (first use in this function)
Common.hsc:242:16: error: 'FF1' undeclared (first use in this function)