| 1 |
Paul de Vrieze posted <200409271113.17794.pauldv@g.o>, excerpted |
| 2 |
below, on Mon, 27 Sep 2004 11:13:17 +0200: |
| 3 |
|
| 4 |
> - Keep your hands of the priority settings that increase priority, when a |
| 5 |
> compilation fails for you it does not mean that it is a blocker. |
| 6 |
|
| 7 |
In this case, the description for "blocker" needs changed. It says |
| 8 |
if it prevents development or use of the app, it's a blocker. Obviously, |
| 9 |
if it doesn't compile, you can't use it, which makes it a blocker by the |
| 10 |
existing definition. |
| 11 |
|
| 12 |
I ran into this just yesterday on a bug I posted |
| 13 |
(http://bugs.gentoo.org/show_bug.cgi?id=65495 ). Currently, the install |
| 14 |
step of the libsdl ebuild won't complete on AMD64, due to lib/lib64 issues |
| 15 |
with a sed fix for a previous bug (34804) specifying lib when the compile |
| 16 |
and rest of the install step uses lib64. It's a fairly new problem as I |
| 17 |
was remerging the same ebuild I'd had merged previously, due to a use |
| 18 |
flag change. While this is indeed a "blocker" on current gentoo-amd64 as |
| 19 |
the bug is certain to occur, and the description said I should /make/ it |
| 20 |
blocker, I debated between major/critical/blocker, and settled for |
| 21 |
critical (as well as assigning to amd64 so it wouldn't bother legacy x86), |
| 22 |
because I recognize the 32-bit majority won't find it "blocker". |
| 23 |
|
| 24 |
Actually, the existing descriptions seem more fitted for a binary |
| 25 |
distribution than from-source, particularly a bug record system primarily |
| 26 |
purposed to catch errors in what are after-all from-source ebuild scripts, |
| 27 |
as the descriptions seem to pre-suppose the errors will be run-time, not |
| 28 |
build-time, in their wording. |
| 29 |
|
| 30 |
* Blocker: Blocks development and/or testing work. |
| 31 |
|
| 32 |
If it can't be merged, that fits, regardless of why. |
| 33 |
|
| 34 |
* Critical: Crashes, loss of data, severe memory leak. |
| 35 |
|
| 36 |
Obvious run-time bias, no guidance about build-time issues at all. (I'd |
| 37 |
certainly rate loss of existing system data while compiling a new app, |
| 38 |
/regardless/ of what it was, blocker, not just critical, but somehow I |
| 39 |
don't think that's what's being described here. =:^# ) |
| 40 |
|
| 41 |
* Major: Major loss of function. |
| 42 |
|
| 43 |
Again, total run-time bias. Logic dictates that if it can't even be |
| 44 |
installed, it /must/ be rated higher than this, since that's |
| 45 |
definitely worse than "major loss of function", because it's loss of ALL |
| 46 |
function! |
| 47 |
|
| 48 |
* Minor: Minor loss of function or other problem where easy workaround is |
| 49 |
present. |
| 50 |
|
| 51 |
Except with packages where upstream /is/ Gentoo, like portage (on which I |
| 52 |
did file a "minor" bug and had it fixed, earlier this year), or where the |
| 53 |
problem is likely upstream anyway, how often could "minor loss of |
| 54 |
function" apply to Gentoo's system /at/ /all/? Granted, the "easy |
| 55 |
workaround" part could logically apply, but if it's blocker by definition |
| 56 |
because the package can't even be installed because the ebuild fails, then |
| 57 |
"easy workaround" or not, it's blocker! |
| 58 |
|
| 59 |
* Trivial: Cosmetic problem like misspelled words or misaligned text. |
| 60 |
|
| 61 |
* Enhancement: Request for enhancement. |
| 62 |
|
| 63 |
OK, these two apply as is. |
| 64 |
|
| 65 |
I'm not so good at proposing concise definitions, but I'd suggest that |
| 66 |
"Minor" be defined in the context of ebuilds as merging OK, but loss of |
| 67 |
some expected functionality. "Major" would be failure to merge but where |
| 68 |
it's likely corner-case (as in depending on a non-standard merge order, |
| 69 |
unusual USE flags, or aggressive CFLAGS) or not widespread. |
| 70 |
|
| 71 |
"Critical" would then be (as I used it) failure to merge cases where it |
| 72 |
applies to most or all of a current arch, possibly excluding |
| 73 |
legacy/majority x86_32 which would be "Blocker". "Blocker" would also |
| 74 |
include bad toolchain failure2merges likely to affect a significant |
| 75 |
amount of people, and common (including arch default) USE or CFLAG |
| 76 |
problems affecting a significant amount of people using an ebuild, |
| 77 |
regardless of arch. |
| 78 |
|
| 79 |
Am I on target here, and should this be bugged under bugzilla, or will the |
| 80 |
discussion result in this being changed on its own so no need? |
| 81 |
|
| 82 |
-- |
| 83 |
Duncan - List replies preferred. No HTML msgs. |
| 84 |
"They that can give up essential liberty to obtain a little |
| 85 |
temporary safety, deserve neither liberty nor safety." -- |
| 86 |
Benjamin Franklin |
| 87 |
|
| 88 |
|
| 89 |
|
| 90 |
-- |
| 91 |
gentoo-dev@g.o mailing list |
Archives