Andrew Wallo
15th December 2008 12:35 UTC

FileRead throws error with *unix style text file and returns no chars.
I'm editing an install of ClamAV. Its configuration is stored in a clamav.conf file which is in unix format. i.e. Only char(10) endings. $\n no $\r$\n's.

ReadFile throws error and quits.
ReadFileBytes throws error and quits.

Is there a macro I can edit that will patch this behavior?

Yes. I've wrapped it in a ifFileExists test. The path is correct.
The call is:

  
ifFileExists "$INSTDIR/conf/clamd.conf" 0 errorNoConfFile  ; if this doesn't exist there is a problem....

       !insertmacro ReplaceInFile "$INSTDIR/conf/clamd.conf" "MailFollowURLs Yes" "MailFollowURLs No"
       ClearErrors
  errorNoConfFile:

Fails.
The ReplaceInFile macro kicks out at FileRead $R0 $R3, which is the first file read line in the macro. I've attempted to modify with FileReadBytes I still can't get data to return from the read. HELP!

I have attached a copy of the file to ask if anyone else can open and successfully read lines from this file. I've had to change the file name to end in .txt for upload.
Thanks.

Animaether
15th December 2008 14:33 UTC

can you read just the first line? perhaps it's trying to read the wrong line (say, past the end of the file or so)?

        FileOpen $0 "c:\temp\clamd.conf" "r"
        FileRead $0 $1
        MessageBox MB_OK "[$1]"
        FileClose $0

Gives...

---------------------------
Test Setup
---------------------------
[##
]
---------------------------
OK   
---------------------------

Using your same code, I got no error... verified that the file is still using only 0A, not 0A0D, and 'MailFollowUrls No' is followed by 0A0A

Andrew Wallo
16th December 2008 20:10 UTC

StepThough.....
Ok, well I confirmed your test. I can acutally open the file seperately, and rip through it. I get all teh ## lines etc. So then I edited the ReplaceInFile's central function called RIP... as you see...I have inserted an output test of the initial parameters, and alerts if any of the preceeding items throw errors. This should enter into its search and hit us with 1000 annoying message boxes as it rips through each line in the file.... however... that doesn't happen. We exit with the first alert, and dump to close because of the errors being set.



What determines the EOF for the FileRead function?

Is there ANY other reason that FileRead would return with the error flag set other than reaching the end of file? Its returning both Empty and Errors set.... indicating the end of file...

But if we manually open the file and read a line we get one.... is there something about being in the macro???

Function RIF

  ClearErrors  ; want to be a newborn

  Exch $0      ; REPLACEMENT
  Exch
  Exch $1      ; SEARCH_TEXT
  Exch 2
  Exch $2      ; SOURCE_FILE

  Push $R0     ; SOURCE_FILE file handle
  Push $R1     ; temporary file handle
  Push $R2     ; unique temporary file name
  Push $R3     ; a line to sar/save
  Push $R4     ; shift puffer
  Push $R5     ; byteHolder
  Push $R6     ; buffer counter
  MessageBox MB_OK "Data before open read. $0 , $1 , $2"
  IfFileExists $2 +1 RIF_error      ; knock-knock
  ifErrors 0 +2
    MessageBox MB_OK "Errors existed before open read. $0 , $1 , $2"
    ClearErrors
  FileOpen $R0 $2 "r"               ; open the door
  ifErrors 0 +2
    MessageBox MB_OK "Errors existed after open read. $2"
    ClearErrors
  GetTempFileName $R2               ; who's new?
  FileOpen $R1 $R2 "w"              ; the escape, please!
  ifErrors 0 +2
    MessageBox MB_OK "Errors existed after open write $2"
    ClearErrors
  RIF_loop:                         ; round'n'round we go
    FileRead $R0 $R3                ; read one line
        MessageBox MB_OK "$R3"

    ifErrors RIF_leaveLoop

    RIF_sar:                        ; sar - search and replace
      Push "$R3"                    ; (hair)stack
      Push "$1"                     ; needle
      Push "$0"                     ; blood
      Call StrReplace               ; do the bartwalk
      StrCpy $R4 "$R3"              ; remember previous state
      Pop $R3                       ; gimme s.th. back in return!
      StrCmp "$R3" "$R4" +1 RIF_sar ; loop, might change again!
    FileWrite $R1 "$R3"             ; save the newbie
  Goto RIF_loop                     ; gimme more

  RIF_leaveloop:                    ; over'n'out, Sir!

    FileClose $R1                   ; S'rry, Ma'am - clos'n now
    FileClose $R0                   ; me 2
    ;MessageBox MB_OK "Files Closed. $R1 , $R0"
    Delete "$2.old"                 ; go away, Sire
    ;MessageBox MB_OK "Deleete $2.old"
    Rename "$2" "$2.old"            ; step aside, Ma'am
    ;MessageBox MB_OK "Rename $2 to $2.old"
    Rename "$R2" "$2"               ; hi, baby!
    ;MessageBox MB_OK "Rename $R2 to $2"
    Delete "$2.old"
    ClearErrors                     ; now i AM a newborn
    Goto RIF_out                    ; out'n'away

  RIF_error:                        ; ups - s.th. went wrong...
    SetErrors                       ; ...so cry, boy!

  RIF_out:                          ; your wardrobe?
  Pop $R6
  Pop $R5
  Pop $R4
  Pop $R3
  Pop $R2
  Pop $R1
  Pop $R0
  Pop $2
  Pop $0
  Pop $1

FunctionEnd

Animaether
16th December 2008 20:20 UTC

well that's the odd thing - even with your new RIF Function with the debug messageboxes - it continues to work just fine. I get a ton of messageboxes and the end result is that in "c:\temp\clamd.conf" (the filename I'm passing), the line "MailFollowURLs Yes" is replaced with "MailFollowURLs No".

Are you saying that it throws up this error..

  ifErrors 0 +2
    MessageBox MB_OK "Errors existed before open read. $0 , $1 , $2"

..?
If so, perhaps include more ClearError / IfErrors -> MessageBox debug points to figure out where on earth it thinks it's erroring. The only time I get that error is if the file does not, in fact, exist at "c:\temp\clamd.conf"

Andrew Wallo
16th December 2008 20:37 UTC

Stupid stupid stupid stupid.
Using forward slashes in file name creates impossible read errors.

Every read after that results in error flag being set. However, NO error flag on FileOpen????? VERY STRANGE.

Andrew Wallo
16th December 2008 20:46 UTC

Thanks for helping me push this to conclusion. I was ( by requirment for ClamAV ) using forward slashed paths because ClamAV needs/expects them in its *nix config file formats. So... my convention read C:/Program Files/clamAV/conf/clamd.conf

Why this doesn't create throws on FileOpen is beyond me.

But converting to C:\\Program Files\clamAV\conf\clamd.conf creates success with the read calls.

???Bug??? FileOpen should throw this error before read consistantly...

Andrew Wallo
16th December 2008 20:56 UTC

And ifFileExists should return an error if that is not a supported path definition.... but ifFileExists would pass the test too!