Hacker News new | ask | show | jobs
by zeotroph 2379 days ago
This is about that ancient EBCDIC (IBM's ASCII "competitor") specific hack called trigraphs, which in modern compilers are opt-in anyhow: "To understand this, I have to admit one thing: I have to pass -trigraphs to a modern version of gcc before this actually works. "

C++17 actually removed support for these.
1 comments
fanf2 2379 days ago
Trigraphs were not specific to EBCDIC: they were also a bad accommodation for the ISO 646 ASCII variants that supported European languages, which used characters {\|} etc. for letters. This is why the newer better header with alternate spellings for restricted character sets is called <iso646.h>

https://en.wikipedia.org/wiki/ISO/IEC_646

link
burfog 2379 days ago
The reasonable accommodation is to substitute with characters that are available in the ISO 646 ASCII variant being used, preferably the characters with the same binary representation as in US ASCII.

Outside of silly games like IOCCC, did anybody actually use trigraphs? I find it really hard to believe that anybody tolerated trigraphs for serious C programming. Was there an obscure European government agency that actually wrote code with trigraphs?

link
boomlinde 2378 days ago
IBM opposed their removal in C++1x

From http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n291...

> The biggest argument against deprecating trigraph is the fact that we know there are source code from world-wide companies, based on an internal survey of our customer problem reports that make heavy use of trigraphs for one reason or another, and customer testimonials. Google code search doesn't find them because these are proprietary code. Further discussions with some of these customers have shown serious dissatisfaction with WG21’s intention to deprecate trigraphs, because they do not wish to make any changes to their source code. It doesn't seem fair to force this group of existing users to change their source code for no apparent benefit, other then reducing an annoyance to another group.

ISO 646 is probably nowhere near as relevant to their argument as EBCDIC is.

> Trigraph deprecation is vexing for existing world-wide companies (of which IBM is one such user) because of their use of EBCDIC variant code pages. It makes it difficult to start in a code page neutral manner, establish what the active code page is, then start using invariant characters. We use a pragma filetag("IBM-1047") to toggle the code page, then we can use #. Prior to that we are in a code page neutral environment and must use the trigraph ??=

link