cmd/cgo: cgo binaries change each time you build them

[alberts]

What steps will reproduce the problem?

cd misc/cgo/stdio
go build hello.go
md5sum hello.go
go build hello.go
md5sum hello.go

What is the expected output?

same hash

What do you see instead?

different hashes

looking at a few changes in the hexdump:

--- hello.hex   2013-03-31 10:33:19.114543157 +0200
+++ hello.1.hex 2013-03-31 10:33:23.269525520 +0200
@@ -41,4 +41,4 @@
-00000280  47 4e 55 00 24 a0 0b 56  b5 4c 4b 86 c0 50 72 55  |GNU.$..V.LK..PrU|
-00000290  44 eb 46 4c f4 d2 c6 af  04 00 00 00 14 00 00 00  |D.FL............|
-000002a0  03 00 00 00 47 4e 55 00  69 a3 c5 7d 98 dc ed 5c  |....GNU.i..}...\|
-000002b0  86 b9 24 f4 92 a5 44 67  a7 6f e5 5a 04 00 00 00  |..$...Dg.o.Z....|
+00000280  47 4e 55 00 13 17 8b 27  51 48 39 83 a0 f6 80 98  |GNU....'QH9.....|
+00000290  09 1f e1 fe 36 ed ab 04  04 00 00 00 14 00 00 00  |....6...........|
+000002a0  03 00 00 00 47 4e 55 00  45 8c 05 ba bf 14 5d 9e  |....GNU.E.....].|
+000002b0  4f c8 d3 f6 d1 6b 52 7f  f6 e1 c9 80 04 00 00 00  |O....kR.........|
@@ -20976,2 +20976,2 @@
-00051f80  04 00 24 00 00 00 08 01  50 00 00 00 01 95 02 00  |..$.....P.......|
-00051f90  00 e2 02 00 00 d0 0f 40  00 00 00 00 00 18 10 40  |.......@.......@|
+00051f80  04 00 24 00 00 00 08 01  50 00 00 00 01 b5 02 00  |..$.....P.......|
+00051f90  00 95 02 00 00 d0 0f 40  00 00 00 00 00 18 10 40  |.......@.......@|
@@ -21049 +21049 @@
-00052410  00 01 c3 03 00 00 e2 02  00 00 20 10 40 00 00 00  |.......... .@...|
+00052410  00 01 a1 03 00 00 95 02  00 00 20 10 40 00 00 00  |.......... .@...|
@@ -21092 +21092 @@
-000526c0  00 00 00 0e b9 03 00 00  01 11 bc 02 00 00 30 10  |..............0.|
+000526c0  00 00 00 0e 07 04 00 00  01 11 bc 02 00 00 30 10  |..............0.|
@@ -21094 +21094 @@
-000526e0  0f a1 03 00 00 01 29 40  10 40 00 00 00 00 00 4e  |......)@[email protected]|
+000526e0  0f ef 03 00 00 01 29 40  10 40 00 00 00 00 00 4e  |......)@[email protected]|
@@ -21628 +21628 @@
-00054840  2d 62 75 69 6c 64 34 35  34 30 36 34 34 36 33 2f  |-build454064463/|
+00054840  2d 62 75 69 6c 64 31 31  35 32 37 37 35 32 37 2f  |-build115277527/|
@@ -21657,7 +21657,7 @@
-00054a10  00 2f 64 65 76 2f 73 68  6d 2f 67 6f 2d 62 75 69  |./dev/shm/go-bui|
-00054a20  6c 64 34 35 34 30 36 34  34 36 33 2f 5f 2f 68 6f  |ld454064463/_/ho|
-00054a30  6d 65 2f 61 6c 62 65 72  74 73 2f 67 6f 2f 6d 69  |me/alberts/go/mi|
-00054a40  73 63 2f 63 67 6f 2f 73  74 64 69 6f 2f 5f 6f 62  |sc/cgo/stdio/_ob|
-00054a50  6a 2f 66 69 6c 65 2e 63  67 6f 32 2e 63 00 2f 68  |j/file.cgo2.c./h|
-00054a60  6f 6d 65 2f 61 6c 62 65  72 74 73 2f 67 6f 2f 6d  |ome/alberts/go/m|
-00054a70  69 73 63 2f 63 67 6f 2f  73 74 64 69 6f 00 5f 6e  |isc/cgo/stdio._n|
+00054a10  00 2f 68 6f 6d 65 2f 61  6c 62 65 72 74 73 2f 67  |./home/alberts/g|
+00054a20  6f 2f 6d 69 73 63 2f 63  67 6f 2f 73 74 64 69 6f  |o/misc/cgo/stdio|
+00054a30  00 2f 64 65 76 2f 73 68  6d 2f 67 6f 2d 62 75 69  |./dev/shm/go-bui|
+00054a40  6c 64 31 31 35 32 37 37  35 32 37 2f 5f 2f 68 6f  |ld115277527/_/ho|
+00054a50  6d 65 2f 61 6c 62 65 72  74 73 2f 67 6f 2f 6d 69  |me/alberts/go/mi|
+00054a60  73 63 2f 63 67 6f 2f 73  74 64 69 6f 2f 5f 6f 62  |sc/cgo/stdio/_ob|
+00054a70  6a 2f 66 69 6c 65 2e 63  67 6f 32 2e 63 00 5f 6e  |j/file.cgo2.c._n|
@@ -21673,8 +21673,8 @@
-00054b10  74 64 6f 75 74 00 73 74  64 65 72 72 00 5f 63 67  |tdout.stderr._cg|
-00054b20  6f 5f 37 36 61 37 61 37  64 61 63 38 66 35 5f 43  |o_76a7a7dac8f5_C|
-00054b30  66 75 6e 63 5f 67 65 74  53 74 64 65 72 72 00 2f  |func_getStderr./|
-00054b40  64 65 76 2f 73 68 6d 2f  67 6f 2d 62 75 69 6c 64  |dev/shm/go-build|
-00054b50  34 35 34 30 36 34 34 36  33 2f 5f 2f 68 6f 6d 65  |454064463/_/home|
-00054b60  2f 61 6c 62 65 72 74 73  2f 67 6f 2f 6d 69 73 63  |/alberts/go/misc|
-00054b70  2f 63 67 6f 2f 73 74 64  69 6f 2f 5f 6f 62 6a 2f  |/cgo/stdio/_obj/|
-00054b80  73 74 64 69 6f 2e 63 67  6f 32 2e 63 00 2f 64 65  |stdio.cgo2.c./de|
+00054b10  74 64 6f 75 74 00 73 74  64 65 72 72 00 2f 64 65  |tdout.stderr./de|
+00054b20  76 2f 73 68 6d 2f 67 6f  2d 62 75 69 6c 64 31 31  |v/shm/go-build11|
+00054b30  35 32 37 37 35 32 37 2f  5f 2f 68 6f 6d 65 2f 61  |5277527/_/home/a|
+00054b40  6c 62 65 72 74 73 2f 67  6f 2f 6d 69 73 63 2f 63  |lberts/go/misc/c|
+00054b50  67 6f 2f 73 74 64 69 6f  2f 5f 6f 62 6a 2f 73 74  |go/stdio/_obj/st|
+00054b60  64 69 6f 2e 63 67 6f 32  2e 63 00 5f 63 67 6f 5f  |dio.cgo2.c._cgo_|
+00054b70  37 36 61 37 61 37 64 61  63 38 66 35 5f 43 66 75  |76a7a7dac8f5_Cfu|
+00054b80  6e 63 5f 67 65 74 53 74  64 65 72 72 00 2f 64 65  |nc_getStderr./de|

Which compiler are you using (5g, 6g, 8g, gccgo)?

6g
gcc-4.7.2-2.fc17.x86_64

Which operating system are you using?

linux
F16 x86-64

Which version are you using?  (run 'go version')

go version devel +35f0e92cbd6c Sat Mar 30 09:22:11 2013 -0700 linux/amd64

Please provide any additional information below.

This was fixed before:

https://code.google.com/p/go/source/detail?r=e24d688078cd

[alberts]

Comment 1:

part of the change might be caused by build ids:
http://fedoraproject.org/wiki/Releases/FeatureBuildId
hopefully the build id is different because of the difference in the cgo bits.

[bradfitz]

Comment 2:

go {build,run} uses temp directories, and those directories get baked in to the binary
in various ways.

[alberts]

Comment 3:

Some discussion:
https://groups.google.com/d/topic/golang-nuts/bjCrKTsr3tE/discussion
As minux noted, we had this feature for a while, but recent changes probably made it go
away.

[rsc]

Comment 4:

Labels changed: added priority-later, go1.2, removed priority-triage.

[robpike]

Comment 5:

Not for 1.2.

Labels changed: removed go1.2.

[rsc]

Comment 6:

Labels changed: added go1.3maybe.

[rsc]

Comment 7:

Labels changed: added release-none, removed go1.3maybe.

[rsc]

Comment 8:

Labels changed: added repo-main.

[bradfitz]

A lot of work has happened on this over the years.

Is this still an issue?

/cc @ianlancetaylor

[ianlancetaylor]

This has been subsumed by #13247. It is fixed for most systems for most compilers, but there are still a few cases remaining. Closing this issue in favor of the one with more information.