> > I just want to compare 2 byte streams in a shell script without using > temporary files.
Just out of curiosity: why? The test directory is already polluted with artefacts during the course of the test-run: I don't see why creating a few more is burdensome. In any case, the difference between GNU od(1) and BSD od(1) is that the latter preserves trailing whitespace: GNU: │0x00000000│ 30 30 30 30 30 30 30 20 20 20 41 20 20 5C 62 20 │0000000 A \b │ │0x00000010│ 20 20 42 20 20 5C 62 20 20 20 43 20 20 20 20 20 │ B \b C │ │0x00000020│ 20 20 44 20 20 5C 6E 0A 30 30 30 30 30 31 30 │ D \n.0000010 │ BSD: │0x00000000│ 30 30 30 30 30 30 30 20 20 20 20 41 20 20 5C 62 │0000000 A \b│ │0x00000010│ 20 20 20 42 20 20 5C 62 20 20 20 43 20 20 20 20 │ B \b C │ │0x00000020│ 20 20 20 44 20 20 5C 6E 20 20 20 20 20 20 20 20 │ D \n │ │0x00000030│ 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 │ │ │0x00000040│ 20 20 20 20 20 20 20 20 0A 30 30 30 30 30 31 30 │ .0000010│ Ergo, simply appending ` *` to your grep(1) expression should be enough to accommodate both flavours. On Sat, 11 Feb 2023 at 12:23, G. Branden Robinson < g.branden.robin...@gmail.com> wrote: > Hi John, > > At 2023-02-11T10:19:08+1100, John Gardner wrote: > > Yes, it builds successfully now (macOS 12.6.3). I ran `make install` > > and the installed Groff seems to be working perfectly. > > Excellent! > > > However, when I ran `make check` in my checkout directory, 7 of the > > tests failed (full details in the attached log file): > [...] > > Are these normal? > > Certainly not. > > > I'm attaching my build logs just in case. > > Thank you--this is helpful. Some or all of these problems I think we > have seen on macOS before; they involve the portability of shell syntax > and of a few (incompletely) standardized tools. xooglers "just want to > serve 5 terabytes". I just want to compare 2 byte streams in a shell > script without using temporary files. Astoundingly, this is way harder > and system-dependent than it should be. I attribute that to the early > mentality that the PDP-11 was the only machine that mattered, and the > belief of the succeeding generation, despite obvious contrary evidence, > that the VAX was the only machine that mattered. > > Nowadays, we laugh at the blinkered folly of our ancestors and > confidently understand that x86-64 is all that matters. > > > FAIL: src/roff/groff/tests/some_escapes_accept_newline_delimiters.sh > > "od -t c". If I remember correctly, macOS od produces output unlike > both Unix Version 7 od and GNU od. Certainly macOS and GNU od don't > agree with each other. > > > FAIL: tmac/tests/an_TS-adds-no-vertical-space.sh > > sed. I'll have to look up what hoop is required for macOS sed, which I > think comes from one of the BSDs but, as I recall, proudly refuses to > admit any means of identifying itself by version number in execution. > (That's for disgusting things like GNU programs.) > > > FAIL: tmac/tests/an_use-input-traps-correctly.sh > > This looks like a shell difference in how sequences of backslashes are > interpreted in double-quoted strings. I see no variable interpolations > in the cases in question, so I may be able to fix this by just making > the strings single-quoted. > > > FAIL: tmac/tests/doc_heading-font-remapping-works.sh > > sed again. It seems to think I am using GNU extensions, but I'm not > using any marked as such in GNU sed's man page. Maybe it's a BRE vs. > ERE thing. POSIX prescribes a -E option for this. I'm sure the sed on > Solaris 10 doesn't support it. :-| > > > FAIL: tmac/tests/latin2_works.sh > > FAIL: tmac/tests/latin5_works.sh > > FAIL: tmac/tests/latin9_works.sh > > "od -t c" again. > > This is useful information, even if it fuels my frustration with > vendors of fundamental Unix utilities. I'll see what I can do. > > Regards, > Branden >
