Re: git-p4: exception when cloning a perforce repository

damien@iwi.me wrote on Thu, 16 Jan 2014 14:46 +0100:

On 16 Jan 2014, at 14:08, Pete Wyckoff [off-list ref] wrote:

quoted

damien@iwi.me wrote on Wed, 15 Jan 2014 09:56 +0100:

quoted

p4 fstat  //depot/openssl/0.9.8j/openssl/include/openssl/bn.h@59702 
... depotFile //depot/openssl/0.9.8j/openssl/include/openssl/bn.h
... headAction edit
... headType symlink
... headTime 1237906419
... headRev 2
... headChange 59702
... headModTime 1231329423

p4 print -q //depot/openssl/0.9.8j/openssl/include/openssl/bn.h#2 | od -c
0000000

p4 print  //depot/openssl/0.9.8j/openssl/include/openssl/bn.h#1        
//depot/openssl/0.9.8j/openssl/include/openssl/bn.h#1 - add change 59574 (text)
p4 print  //depot/openssl/0.9.8j/openssl/include/openssl/bn.h#2
//depot/openssl/0.9.8j/openssl/include/openssl/bn.h#2 - edit change 59702 (symlink)

That's interesting.  When I do the equivalent "p4 print" commands
it shows something like this.

arf-git-test$ p4 fstat //depot/bn.h
... depotFile //depot/bn.h
... clientFile /dev/shm/trash directory.t9802-git-p4-filetype/cli/bn.h
... isMapped 
... headAction edit
... headType symlink
... headTime 1389876870
... headRev 2
... headChange 8
... headModTime 1389876870
... haveRev 2

arf-git-test$ p4 print //depot/bn.h#1
//depot/bn.h#1 - add change 7 (text)
file-text

arf-git-test$ p4 print //depot/bn.h#2
//depot/bn.h#2 - edit change 8 (symlink)
/elsewhere/bn.h

I don't know how you manage to get a symlink with an empty
destination like that.

I'll work on a way to hack around this failure.  In the mean time,
if you're game, it might be fun to see what p4 does with such a
repository.  You could make a client for just that little subdir,
check out at 59702 and see what is there:

mkdir testmess
cd testmess
cat <<EOF | p4 client -i
Client: testmess
Description: testmess
Root: $(pwd)
View: //depot/openssl/0.9.8j/openssl/include/openssl/... //testmess/...
EOF

then take a look at how p4 represents the "empty" symlink
in the filesystem:

p4 sync @59702
ls -la bn.h

I’ve tried exactly your commands, and I’ve got an empty folder..

{14:38}~/p4/testmess ➭ p4 sync @59702
//depot/openssl/0.9.8j/openssl/include/openssl/aes.h#2 - refreshing /home/dgerard/p4/testmess/aes.h
//depot/openssl/0.9.8j/openssl/include/openssl/asn1.h#2 - refreshing /home/dgerard/p4/testmess/asn1.h
//depot/openssl/0.9.8j/openssl/include/openssl/asn1_mac.h#2 - refreshing /home/dgerard/p4/testmess/asn1_mac.h
//depot/openssl/0.9.8j/openssl/include/openssl/asn1t.h#2 - refreshing /home/dgerard/p4/testmess/asn1t.h
//depot/openssl/0.9.8j/openssl/include/openssl/bio.h#2 - refreshing /home/dgerard/p4/testmess/bio.h
//depot/openssl/0.9.8j/openssl/include/openssl/blowfish.h#2 - refreshing /home/dgerard/p4/testmess/blowfish.h
//depot/openssl/0.9.8j/openssl/include/openssl/bn.h#2 - refreshing /home/dgerard/p4/testmess/bn.h
//depot/openssl/0.9.8j/openssl/include/openssl/buffer.h#2 - refreshing /home/dgerard/p4/testmess/buffer.h
[…]


{14:39}~/p4/testmess ➭ ls -la
total 12
drwxr-xr-x 2 dgerard dgerard 4096 janv. 16 14:37 .
drwxr-xr-x 4 dgerard dgerard 4096 janv. 16 14:34 ..
-rw-r--r-- 1 dgerard dgerard   93 janv. 16 14:37 .perforce	


Then I tried to sync the previous changeset, which is ok :

{14:44}~/p4/testmess ➭ p4 sync -f @59701
//depot/openssl/0.9.8j/openssl/include/openssl/aes.h#1 - updating /home/dgerard/p4/testmess/aes.h
[…]

{14:44}~/p4/testmess ➭ l
total 0
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 aes.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 asn1.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 asn1_mac.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 asn1t.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 bio.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 blowfish.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 bn.h
-r--r--r-- 1 dgerard dgerard 0 janv. 16 14:44 buffer.h
[…]



However, when trying to sync to the appropriate changeset :

{14:44}~/p4/testmess ➭ p4 sync -f @59702
//depot/openssl/0.9.8j/openssl/include/openssl/aes.h#2 - updating /home/dgerard/p4/testmess/aes.h
rename: /home/dgerard/p4/testmess/aes.h: No such file or directory
//depot/openssl/0.9.8j/openssl/include/openssl/asn1.h#2 - updating /home/dgerard/p4/testmess/asn1.h
rename: /home/dgerard/p4/testmess/asn1.h: No such file or directory
//depot/openssl/0.9.8j/openssl/include/openssl/asn1_mac.h#2 - updating /home/dgerard/p4/testmess/asn1_mac.h
rename: /home/dgerard/p4/testmess/asn1_mac.h: No such file or directory
//depot/openssl/0.9.8j/openssl/include/openssl/asn1t.h#2 - updating /home/dgerard/p4/testmess/asn1t.h
rename: /home/dgerard/p4/testmess/asn1t.h: No such file or directory
//depot/openssl/0.9.8j/openssl/include/openssl/bio.h#2 - updating /home/dgerard/p4/testmess/bio.h
rename: /home/dgerard/p4/testmess/bio.h: No such file or directory
//depot/openssl/0.9.8j/openssl/include/openssl/blowfish.h#2 - updating /home/dgerard/p4/testmess/blowfish.h
rename: /home/dgerard/p4/testmess/blowfish.h: No such file or directory
//depot/openssl/0.9.8j/openssl/include/openssl/bn.h#2 - updating /home/dgerard/p4/testmess/bn.h
rename: /home/dgerard/p4/testmess/bn.h: No such file or directory

And the folder remains untouched.
Quite strange for me...

Oh cool, that helps a lot.  P4 is just broken here, so we can get
away with being a bit sloppy in git.  I'll try just pretending
"empty symlinks" are not in the repo.  Hopefully you'll have a
future commit in your p4 repo that brings back bn.h properly.

Still not sure about how I'll test this.

Thanks,

		-- Pete