On Mon, 12 Jul 2021 13:48:52 -0500, Kirk Wolf wrote: > ># a "here" document with 10,000 60-byte lines: > ... >FSUM9225 no memory: EDC5132I Not enough memory. (errno2=0xC5070003) > ... >We took a guess that realloc() was the problem with the shell, and sure >enough if you set this envar *before* running the shell, then the here >document works as expected: > >export _CEE_REALLOC_CONTROL=32K,25 > Grrr... I had understood that (implementation dependend) both here-documents and command-substitutions wrote to temp files and read them back. In consequence: o stat() guides malloc(); no need for realloc(). o you exhaust file space instead.
What does bash do? (No solace to an ISV, of course.) There are always implementation limits. They should be documented; you shouldn't need to "take a guess" or even resort to "User Response" in M&C. IBM has reacted to such SRs at times with "Do you really need that to work, or are you just testing? What's the business case for you and for IBM?" (I test; doesn't IBM?) -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
