Issues 2008-08-04
 Share
The version of the browser you are using is no longer supported. Please upgrade to a supported browser.Dismiss

 
View only
 
 
Still loading...
ABCDEFG
1
IDVotesType/MemberMilestonePriOwnerSummary/Comments
2
3
21243DEFECTtriageP3stevenknightInstall() race condition
4
consensus
5
6
Bill
7
Brandon1.xReproduced the problem and updated the bug.
8
David
9
Gary
10
GreginvalidClearly a problem with their installation or configuration; not a bug in SCons itself.
11
Jim
12
Ken
13
Sohail
14
Steven1.x ?p3 ?stevenknight ?I'm not sure about invalid; we've had problems with open file handles causing problems on Windows in the past. I'd love to just tell people to not use Windows, but having parallel SCons not work reliably at all for software on the most popular OS just isn't viable. His theory about shutil.copy2() not using our override that disables the inherited filehandles is good. shutil.copy2() isn't doing anything magic, so just using our own implementation would probably be an improvement.
15
16
21550DEFECTtriageP4stevenknightGetBuildPath and File.path discrepancy
17
consensus
18
19
Bill
20
BrandoninvalidDrop a @deprecated or equivalent on the function?
21
David
22
Gary
23
Greginvalid or depends 2126One is a path to the virtual file, one is a path to the real file. To my surprise, GetBuildPath() is the one that shows the "real" path (it should be named GetSourcePath() if it's going to do that). Moreover, GetBuildPath() returns the name relative to 'build' (hence, the full path name since it's not under 'build') and f.path returns the path relative to the top directory.

He's right; these are very confusing and inconsistent, but is it a bug? The man page is no help; it could be describing almost anything.

The logical thing to do would be to deprecate GetBuildPath in favor of Node.bldpath (or some such); that is, as an Node virtual attribute, one of the ones to be implemented by issue 2126. In that case, we could make this issue dependent on 2126 and have it deprecate GetBuildPath.

(On the other hand, if this example works at all, the man page updates I just did about variant_dir= are wrong.)
24
Jim
25
Ken
26
Sohail
27
SteveninvalidDeprecating GetBuildPath()++. I can't even remember why it was added, and have no idea if it's really used much. It may even pre-date the addition of BuildDir/VariantDir, which might account for the inconsistent use of "Build" in the name. From looking at the code, the only that makes sense to me is that it was an attempt to give users access to the functionality in arg2nodes(), so you could pass th string to a public API (with capital letters and everything!) and get back the same Node that a call to a Builder would return.
28
29
21570DEFECTtriageP4stevenknightscons cannot reliably be stopped by ctr+C in configure contexts
30
consensus
31
32
Bill
33
Brandon1.xOn Windows Ctrl+C is not guaranteed to kill sub processes so if you try to run again its fairly common to see an old process stuck waiting for stdin with file handles open. Not sure if that is related to this bug.
34
David
35
Gary
36
GregI hate timing bugsI'm not so sure his test is valid; it needs to be a _lot_ longer. Even if ^C is typed the instant the output starts, the standard *IX write buffering would give the command plenty of time to finish long before the signal is received. We need to determine both that the signal is actually received during processing and exactly where it is received.
37
Jim
38
Ken
39
Sohail
40
Steven1.xp3?The "fail and not re-run" part is a definite bug. I'd love it if this could just be punted until we come up with a better mechanism entirely, but that's probably not realistic. Right now, I'm inclined to give it my stock, "re-prioritize in the 1.x timeframe" category... :-(
41
42
21580PATCHtriageP4stevenknightfetch_win32_parallel_msg doesn't print message
43
consensus
44
45
Bill
46
Brandon
47
David
48
Gary
49
Greg1.0.1p4anybodyThe patch is incorrect (it should only print something if the text is not None), but that's trivial to fix. Is there a way to test it?
50
Jim
51
Ken
52
Sohail
53
Steven1.0.1p4anybodyProbably testable via monkey-patching the library modules. Not sure if it's worth an automated test, though.
54
Loading...
 
 
 
Sheet 1