AIX: revert xsetprogname and xgetprogname #4056
Conversation
Turns out some platforms (such as AIX) don't implement these. Don't compile it just yet, it needs some fixing up, which will be done in the next commit. Related: rpm-software-management#4053 This reverts commit fa06b68.
Fix minor issues with the original code reverted in the previous commit that prevented it from compiling cleanly (mostly around the const usage) and actually enable it if no system implementation is available (such as on AIX). Note that this new source file is tiny so it's included in libmisc instead of a new target, and librpm is linked against it. Related: rpm-software-management#4053
|
Note that it's split into two commits (revert of the original + fixups) to make reviewing easier. I can squash it afterwards (or we can leave it this way if you prefer so). |
|
Hmm. The idea of reviving the original assumed the original implementation made sense, but it doesn't, really. Which must've contributed to it getting deleted. Thing is, this is an incredibly fiddly and fragile tangle of ifdefs and weird symbols scattered around, and impossible to test outside some specific platform that doesn't match the previous tangle of stuff. The sane way to do such portability wrappers is to have proper functions that are always compiled on all platforms, and then handle all the platform specifics inside that one source file, and provide them in the public API. For rpm this would mean something like rgetprogname() and rsetprogname() in librpmio. include/rpm/rpmutil.h is a fine place for these. And none of these to-link-or-not-to-link and weird _rpmxgrubfargh stuff 😄 It'll also make system.h smaller, which is always a good sign. |
|
Yep, I did test this by removing the cmake & header conditionals but otherwise agree, it's kinda messy this way. The precedent where we use our own r*() wrappers makes sense, indeed. Let's try that, then 👍 |
|
FWIW, my recollection of the original implementation was that it did something like that, but it was far messier than my memory (a rare occurence, that 😆 ) |
2 participants
Another take at #4053.