Error in gzfile(file, mode) when checking a package with rcmd check

Following Uwe Ligges' instructions (off list) I have updated Duncan
Murdoch's toolchain, with no effect. The trouble was solved when I moved
the package to a directory with a shorter path (U:\Documents and
Settings\pgiraudo\Bureau instead of U:\Documents and
Settings\pgiraudo\Mes documents\R\pgir_arch\pgirmess_arch\On work) and
then things went trough smoothly.

What puzzles me is that the "long" path did not make a problem until R
2.13.1 (this is the directory in which I worked since years for package
development) and that something seems to have happened after upgrading
to R 2.13.2

Anyway. The main thing: I have a way to package now.

Patrick







Le 16/10/2011 17:54, Joshua Wiley a ?crit :

Hmm, I have not had any difficulty with 2.13.2 on
x86_64-pc-mingw32/x64, but I do have trouble with R CMD check on R
devel, as has another user now. It is never with building or
installing, just the checking. In my case, the error comes when
building the manual.

2011/10/16 Patrick Giraudoux<patrick.giraudoux at univ-fcomte.fr>:

Also have just tried to build the package with no preliminary check. The
package apparently builds OK, tar.gz as well as binary .zip. Really
looks
like the error has to do with the rcmd check command specifically...
May it
be a bug ? Suppose I am not the only one to have checked a package
since R
2.13.2 has been delivered ?

Patrick


Le 16/10/2011 11:02, Patrick Giraudoux a ?crit :

PS: looks like if it has something to do with moving from R 2.13.1 to R
2.13.2, since I cannot even check earlier versions of the pgirmess
package
that where well checked and compiled under earlier R versions.


Le 16/10/2011 09:50, Patrick Giraudoux a ?crit :

Hi,

For the first time I have a strange behaviour when checking a package
before 'packaging' the code. Looks like a file cannot be read.

rcmd check pgirmess

* using log directory 'U:/Documents and Settings/pgiraudo/Mes
documents/R/pgir_arch/pgirmess_arch/On work/pgirmess.Rcheck'
* using R version 2.13.2 (2011-09-30)
* using platform: i386-pc-mingw32 (32-bit)
* using session charset: ISO8859-1
* checking for file 'pgirmess/DESCRIPTION' ... OK
* this is package 'pgirmess' version '1.5.2'
* checking package dependencies ... OK
* checking if this is a source package ... OK
* checking for .dll and .exe files ... OK
* checking whether package 'pgirmess' can be installed ... ERROR
Installation failed.
See 'U:/Documents and Settings/pgiraudo/Mes
documents/R/pgir_arch/pgirmess_arch/On
work/pgirmess.Rcheck/00install.out'
for details.


In '00install.out', I get:

* installing *source* package 'pgirmess' ...
** R
** data
** preparing package for lazy loading
** help
*** installing help indices
** building package indices ...
** testing if installed package can be loaded Error in gzfile(file,
mode)
: cannot open the connection
Calls: %in% -> match -> installed.packages -> saveRDS -> gzfile
In addition: Warning message:
In gzfile(file, mode) :
cannot open compressed file
'U:\DOCUME~1\Admin\LOCALS~1\Temp\RtmpwWcmem/libloc_U%3a%2fDocuments%20and%20Settings%2fpgiraudo%2fMes%20documents%2fR%2fpgir_arch%2fpgirmess_arch%2fOn%20work%2fpgirmess.RcheckVersion,Priority,Depends,Imports,LinkingTo,Suggests,Enhances,OS_type,License,Archs,Built.rds',

probable reason 'No such file or directory'
Execution halted
ERROR: loading failed
* removing
'U:/DOCUME~1/pgiraudo/MESDOC~1/R/PGIR_A~1/PGIRME~1/ONWORK~1/PGIRME~1.RCH/pgirmess'


Any hint welcome ! I am stuck on that.

Patrick

On 06.11.2011 18:46, Patrick Giraudoux wrote:

Following Uwe Ligges' instructions (off list) I have updated Duncan
Murdoch's toolchain, with no effect. The trouble was solved when I moved
the package to a directory with a shorter path (U:\Documents and
Settings\pgiraudo\Bureau instead of U:\Documents and
Settings\pgiraudo\Mes documents\R\pgir_arch\pgirmess_arch\On work) and
then things went trough smoothly.

It is really "Mes documents" or just what Microsoft translates this 
way in the Windows Explorer? It may help to use the correct names then.

Anyway, I have seen another problem where non ASCII characters matter 
and will digg...

Best,
Uwe Ligges

What puzzles me is that the "long" path did not make a problem until R
2.13.1 (this is the directory in which I worked since years for package
development) and that something seems to have happened after upgrading
to R 2.13.2

Anyway. The main thing: I have a way to package now.

Patrick







Le 16/10/2011 17:54, Joshua Wiley a ?crit :

Hmm, I have not had any difficulty with 2.13.2 on
x86_64-pc-mingw32/x64, but I do have trouble with R CMD check on R
devel, as has another user now. It is never with building or
installing, just the checking. In my case, the error comes when
building the manual.

2011/10/16 Patrick Giraudoux<patrick.giraudoux at univ-fcomte.fr>:

Also have just tried to build the package with no preliminary 
check. The
package apparently builds OK, tar.gz as well as binary .zip. Really
looks
like the error has to do with the rcmd check command specifically...
May it
be a bug ? Suppose I am not the only one to have checked a package
since R
2.13.2 has been delivered ?

Patrick


Le 16/10/2011 11:02, Patrick Giraudoux a ?crit :

PS: looks like if it has something to do with moving from R 2.13.1 
to R
2.13.2, since I cannot even check earlier versions of the pgirmess
package
that where well checked and compiled under earlier R versions.


Le 16/10/2011 09:50, Patrick Giraudoux a ?crit :

Hi,

For the first time I have a strange behaviour when checking a 
package
before 'packaging' the code. Looks like a file cannot be read.

rcmd check pgirmess

* using log directory 'U:/Documents and Settings/pgiraudo/Mes
documents/R/pgir_arch/pgirmess_arch/On work/pgirmess.Rcheck'
* using R version 2.13.2 (2011-09-30)
* using platform: i386-pc-mingw32 (32-bit)
* using session charset: ISO8859-1
* checking for file 'pgirmess/DESCRIPTION' ... OK
* this is package 'pgirmess' version '1.5.2'
* checking package dependencies ... OK
* checking if this is a source package ... OK
* checking for .dll and .exe files ... OK
* checking whether package 'pgirmess' can be installed ... ERROR
Installation failed.
See 'U:/Documents and Settings/pgiraudo/Mes
documents/R/pgir_arch/pgirmess_arch/On
work/pgirmess.Rcheck/00install.out'
for details.


In '00install.out', I get:

* installing *source* package 'pgirmess' ...
** R
** data
** preparing package for lazy loading
** help
*** installing help indices
** building package indices ...
** testing if installed package can be loaded Error in gzfile(file,
mode)
: cannot open the connection
Calls: %in% -> match -> installed.packages -> saveRDS -> gzfile
In addition: Warning message:
In gzfile(file, mode) :
cannot open compressed file
'U:\DOCUME~1\Admin\LOCALS~1\Temp\RtmpwWcmem/libloc_U%3a%2fDocuments%20and%20Settings%2fpgiraudo%2fMes%20documents%2fR%2fpgir_arch%2fpgirmess_arch%2fOn%20work%2fpgirmess.RcheckVersion,Priority,Depends,Imports,LinkingTo,Suggests,Enhances,OS_type,License,Archs,Built.rds', 


probable reason 'No such file or directory'
Execution halted
ERROR: loading failed
* removing
'U:/DOCUME~1/pgiraudo/MESDOC~1/R/PGIR_A~1/PGIRME~1/ONWORK~1/PGIRME~1.RCH/pgirmess' 



Any hint welcome ! I am stuck on that.

Patrick

Means "Mes documents" seems not to be the real cause, but the "On 
Work" directory. However, replacing "On Work" by "On_work" gives a 
failure as before. Conclusion: looks like if the "space" character is 
not the problem, but the length of the path or the number of embedded 
directories (?).

Le dimanche 06 novembre 2011 ? 19:31 +0100, Patrick Giraudoux a ?crit :

Means "Mes documents" seems not to be the real cause, but the "On
Work" directory. However, replacing "On Work" by "On_work" gives a
failure as before. Conclusion: looks like if the "space" character is
not the problem, but the length of the path or the number of embedded
directories (?).

Is "On Work" the actual name, or a translation? Accentuated characters
are often the cause of bugs in filenames.


Cheers

Le 06/11/2011 19:47, Milan Bouchet-Valat a ?crit :

Le dimanche 06 novembre 2011 ? 19:31 +0100, Patrick Giraudoux a ?crit :

Means "Mes documents" seems not to be the real cause, but the "On
Work" directory. However, replacing "On Work" by "On_work" gives a
failure as before. Conclusion: looks like if the "space" character is
not the problem, but the length of the path or the number of embedded
directories (?).

Is "On Work" the actual name, or a translation? Accentuated characters
are often the cause of bugs in filenames.


Cheers

Yes. There is no accent just a space as in "Documents and settings" or
"My Documents" or "Mes documents". To check if there was really a
'space' problem, I have renamed "On Work" to "On_work" (with no space),
and rcmd check does not work better.

Rcmd check works fine only when the pgirmess folder is moved one
directory up (so still on a path that still have 'space' in directory
names, e.g. "Documents and Settings" and "Mes documents".

Quite strange.

On 11-11-06 1:59 PM, Patrick Giraudoux wrote:

Le 06/11/2011 19:47, Milan Bouchet-Valat a ?crit :

Le dimanche 06 novembre 2011 ? 19:31 +0100, Patrick Giraudoux a ?crit :

Means "Mes documents" seems not to be the real cause, but the "On
Work" directory. However, replacing "On Work" by "On_work" gives a
failure as before. Conclusion: looks like if the "space" character is
not the problem, but the length of the path or the number of embedded
directories (?).

Is "On Work" the actual name, or a translation? Accentuated characters
are often the cause of bugs in filenames.


Cheers

Yes. There is no accent just a space as in "Documents and settings" or
"My Documents" or "Mes documents". To check if there was really a
'space' problem, I have renamed "On Work" to "On_work" (with no space),
and rcmd check does not work better.

Rcmd check works fine only when the pgirmess folder is moved one
directory up (so still on a path that still have 'space' in directory
names, e.g. "Documents and Settings" and "Mes documents".

Quite strange.

Windows historically had quite a short limit on pathnames (something like
260 characters, or bytes), and also on command lines. ?I don't know what the
current limits are, but R has the 260 limit coded into it, and likely some
of our other tools do too. ?Your paths don't approach that limit, but once
you put several copies of them into one command line, you would. ?I suspect
that's what is catching you.

Duncan Murdoch